WO2016155853A1

WO2016155853A1 - Multii-band signal compressing

Info

Publication number: WO2016155853A1
Application number: PCT/EP2015/078536
Authority: WO
Inventors: Genaro Wölfl; Florian Wolf
Original assignee: Harman Becker Automotive Systems Gmbh
Priority date: 2015-04-02
Filing date: 2015-12-03
Publication date: 2016-10-06

Abstract

An audio signal compressor is configured to compress by amplitude an input audio signal within an audio signal frequency band to provide an output audio signal, the compressor comprising at least two compression frequency bands in which the audio signal is compressed, at least one of the compression frequency bands being narrower than the complete audio signal frequency band.

Description

MULTI-BAND SIGNAL COMPRESSING

BACKGROUND

1. Technical Field.

[0001] The disclosure relates to a system and method (generally referred to as a "system") for processing a signal.

2. Related Art.

[0002] Protection limiters for loudspeakers are used in several applications, including low power audio systems such as mobile phones, tablets, laptops and portable loudspeaker systems, and high power systems such as active loudspeakers for home audio, studio monitors, car audio systems and public address (PA) systems. The loudspeakers used in these applications range from miniature loudspeakers of less than 1cm diameter in mobile devices to huge bass loudspeakers measuring more than 50cm in PA systems. Nevertheless, the available sound pressure level (SPL) is in most cases limited by the loudspeaker and the application would benefit from increased SPL capability. Protection limiters will generally limit the available SPL to various extents. Furthermore, the limitation process will induce various undesired audible artifacts, such as noticeably varying loudness (pumping effect), tonality changes and uncorrelated noise.

SUMMARY [0003] An audio signal compressor is configured to compress by amplitude an input audio signal within an audio signal frequency band to provide an output audio signal, the compressor comprising at least two compression frequency bands in which the audio signal is compressed, at least one of the compression frequency bands being narrower than the complete audio signal frequency band.

[0004] An audio signal compressing method is configured to compress by amplitude an input audio signal within an audio signal frequency band to provide an output audio signal, the compressor comprising at least two compression frequency bands in which the audio signal is compressed, at least one of the compression frequency bands being narrower than the complete audio signal frequency band.

[0005] Other systems, methods, features and advantages will be, or will become, apparent to one with skill in the art upon examination of the following figures and detailed description. It is intended that all such additional systems, methods, features and advantages be included within this description, be within the scope of the invention, and be protected by the following claims.

BRIEF DESCRIPTION OF THE DRAWINGS

[0006] The system may be better understood with reference to the following drawings and description. The components in the figures are not necessarily to scale, emphasis instead being placed upon illustrating the principles of the invention. Moreover, in the figures, like referenced numerals designate corresponding parts throughout the different views.

[0007] Figure 1 is a block diagram of an exemplary application of a single channel multi- band compressor for the protection of a single loudspeaker.

[0008] Figure 2 is a block diagram of another exemplary application of a single channel multi-band compressor in connection with at least one additional compressor for the protection of a single loudspeaker.

[0009] Figure 3 is a block diagram of another exemplary application of a single channel multi-band compressor for protection of multiple loudspeakers which are interconnected by a passive crossover network.

[0010] Figure 4 is a block diagram of another exemplary application of a single channel multi-band compressor for protection of multiple loudspeakers which are interconnected by a multi-amplifier network.

[0011] Figure 5 is a signal flow diagram illustrating the structure of an exemplary single channel multi-band compressor module with serial signal flow and a multiplicity of fractional-bands.

[0012] Figure 6 is a signal flow diagram illustrating the structure of an exemplary single channel multi-band compressor module with parallel signal flow and a multiplicity of fractional-bands. [0013] Figure 7 is a signal flow diagram illustrating the structure of an exemplary single channel multi-band compressor module with parallel signal flow and a single fractional-band.

[0014] Figure 8 is diagram illustrating equal loudness curves according ISO standard.

[0015] Figure 9 is a diagram illustrating loudness curves referenced to the 80 phon curve.

[0016] Figure 10 is a diagram illustrating the loudness curves shown in Figure 9, wherein the curves above the 80 phon curves are shifted downwards by their nominal sound pressure level increase.

[0017] Figure 11 is a magnitude-versus-frequency diagram illustrating an exemplary target curve approximated with ten infinite impulse response filter biquads.

[0018] Figure 12 is an amplitude-over- frequency diagram illustrating the amplitude- frequency responses of various band filters.

[0019] Figure 13 is an amplitude-over-frequency diagram illustrating exemplary attenuations versus frequency as applied in fractional-bands with frequency response below approximately 200Hz and above approximately 10kHz, determined by equal loudness curves.

[0020] Figure 14 is an amplitude-over- frequency diagram illustrating exemplary amplitude- frequency responses of 12 parametric filters with parameter maps set to track equal loudness limits.

[0021] Figure 15 is an amplitude-over- frequency diagram illustrating exemplary frequency responses of 12 parametric filters with parameter maps set to track equal loudness limits and with matching for multiple limitation levels.

[0022] Figure 16 is a signal flow diagram illustrating the structure of an exemplary single channel multi-band compressor module with multiple compressor blocks per frequency band.

[0023] Figure 17 is a signal flow diagram illustrating the structure of an exemplary single channel multi-band compressor module with multiple target curves per frequency band.

[0024] Figure 18 is a signal flow diagram illustrating the structure of an exemplary multi- band compressor with multiple parallel filters in the main signal path.

[0025] Figure 19 is a signal flow diagram illustrating the structure of another exemplary multi-band compressor with multiple parallel filters in the main signal path.

[0026] Figure 20 is a diagram illustrating loudness curves below the 80 phon curve, which are referenced to the 80 phon curve and shifted upwards by their nominal SPL decrease.

[0027] Figure 21 is a signal flow diagram illustrating the structure of another exemplary multi-band limiter in which higher harmonics for virtual bass signal generation are added, the higher harmonics being generated by a phase vocoder based on the attenuation of the fundamental frequency.

[0028] Figure 22 is a signal flow diagram illustrating the structure of another exemplary multi-band limiter in which filtering of the audio signal input into a harmonics generator determines the fundamental frequency range for virtual bass signal generation.

[0029] Figure 23 is a block diagram illustrating a multi-channel compressor with multiple multi-band compressor modules and a multi-channel link module.

[0030] Figure 24 is a signal flow diagram illustrating the structure of an exemplary multi- band compressor which is adapted for use with the multi-channel link module shown in Figure 23.

[0031] Figure 25 is a signal flow diagram illustrating the structure of an exemplary multichannel link module with multiple linking groups and a group linker applicable in the multichannel compressor shown in Figure 23.

[0032] Figure 26 is a signal flow diagram illustrating the structure of an exemplary multi- band compressor with current evaluation, which is configured to limit the current supplied to a loudspeaker.

[0033] Figure 27 is a signal flow diagram illustrating the structure of an exemplary multi- band compressor with power evaluation, which is configured to limit power consumed by a loudspeaker.

[0034] Figure 28 is a signal flow diagram illustrating a structure in which a general structure thermal limiter is combined with a multi-band limiter.

[0035] Figure 29 is a signal flow diagram illustrating a structure in which a thermal limiter is integrated into a multi-band compressor module with serial signal flow.

[0036] Figure 30 is a signal flow diagram illustrating a structure in which a thermal limiter is integrated into a multi-band compressor module with parallel signal flow.

[0037] Figure 31 is a signal flow diagram of a structure for calculating the power supplied to a loudspeaker based on the applied voltage and its DC resistance.

[0038] Figure 32 is a block diagram of an exemplary application of a multi-band compressor module with integrated thermal limiter.

[0039] Figure 33 is a diagram illustrating the impedance and admittance over frequency of an exemplary loudspeaker. [0040] Figure 34 is a signal flow diagram of a structure for calculating the power supplied to a loudspeaker based on the applied voltage and the loudspeaker's alternating current (AC) impedance and direct current (DC) resistance.

[0041] Figure 35 is a diagram illustrating the magnitude response of an admittance filter based on the admittance shown in Figure 33.

[0042] Figure 36 is a circuit diagram illustrating a thermal model by analogy with electrical circuits.

[0043] Figure 37 is a signal flow diagram illustrating a thermal limiter for loudspeakers with passive crossover filters in connection with a multi-band compressor module with serial signal flow.

[0044] Figure 38 is a signal flow diagram illustrating the structure of a multi-band compressor module with central gain distribution unit.

[0045] Figure 39 is a signal flow diagram illustrating the structure of the central gain distribution unit used in the multi-band compressor module shown in Figure 38.

[0046] Figure 40 is a signal flow diagram illustrating the structure of the central gain distribution unit used in the multi-band compressor module of Figure 38 with full-band gain absorbed into fractional-bands.

[0047] Figure 41 is a signal flow diagram illustrating the structure of a multi-band compressor module with attenuation calculation in nested feedback loops.

[0048] Figure 42 is a signal flow diagram illustrating the structure of a multi-band compressor module with attenuation calculation in separate feedback loops.

[0049] Figure 43 is a signal flow diagram illustrating the structure of a multi-band compressor module with attenuation calculation in separate feedback loops with variable gain element. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0050] A compressor may be understood as a system or method which for a given amplitude range of input signals produces a smaller amplitude range of output signals. This may be accomplished by amplifying weak signals and/or attenuating strong signals. A limiter may be understood as a system or method in which the output is automatically prevented from exceeding a predetermined value. For example, the output amplitude may be substantially linear with regard to the input amplitude up to the predetermined value and substantially constant thereafter. For signals having both positive and negative values, the predetermined value is usually independent of signal. For the sake of simplicity, the term "compressor" is used herein for both compressors and limiters or a blend of them unless expressly distinguished.

[0051] Figure 1 shows an exemplary single channel multi-band compressor 101 which may drive via a power amplifier 102 a loudspeaker 103 which should be protected. In any case, a positive gain at any intermittent stage (e.g., an optional digital to analog converter 104 if the compressor 101 is implemented in the digital domain) between the compressor 101 and the loudspeaker 103 should be taken into account in the compressor 101. Negative gain (also referred to as attenuation) downstream of the compressor 101 may lead to excessive limitation but will not harm the loudspeaker 103. At least one additional compressor 201 may be inserted downstream (or upstream) of the compressor 101, as shown in Figure 2, in order to protect the loudspeaker 103, e.g., from overheating. This is only required if there is no such protection implemented in the compressor 101.

[0052] The single channel multi-band compressor 101 may protect a single loudspeaker 103, as shown in Figures 1 and 2, or multiple loudspeakers 301 and 302, connected in parallel or series with or without an optional passive crossover network 303 between loudspeakers 301 and 302 and amplifier 102 as shown in Figure 3. The optional crossover network 303 may be of the parallel or series type.

[0053] Another exemplary application of the single channel multi-band compressor 101 may include amplifier-loudspeaker combinations driven with the same signal as shown in Figure 4. The two exemplary combinations shown in Figure 4 include amplifiers 401, 402 and loudspeakers 403, 404. No inherent constraints exist with regard to the type of loudspeaker to be protected. Protection is possible for, but not limited to, dynamic loudspeakers, electro- dynamic loudspeakers, horn loudspeakers, piezoelectric speakers, magneto strictive speakers, electrostatic loudspeakers, ribbon and planar magnetic loudspeakers, bending wave loudspeakers, flat panel loudspeakers, air motion transducers, plasma arc speakers, digital speakers, transparent ionic conduction speakers, thermo-acoustic speakers or speakers based on dielectric elastomers. The compressor modules described below may be generally applied in all fields of audio signal limiting and compressing. [0054] In a first example, the compressor 101 may be provided as a single channel multi- band compressor module 501 with a serial-signal-flow structure as shown in Figure 5. A structure, a module and their components may include at least one of hardware, software and signal flow. Digital input audio data x(n), also referred to as input signal x(n) or audio signal x(n), are input to the compressor module 501 and the processed audio data are output as digital output audio data y(n), also referred to as output signal y(n) or amplitude limited/boosted (in the following only referred to as compressed) output signal y(n). Compressor module 501 combines a fractional-band compressor section 502 with a full-band compressor section 503. Both parts of the compressor module 501, i.e., the fractional-band compressor section 502 and the full-band compressor section 503, deliver in combination a better performance than either part would do if used alone.

[0055] Weighting as used herein may be performed in discrete steps or continuously over frequency in accordance with a weighting function. If a certain target is to be achieved, weighting is performed according to a corresponding target function, e.g., the transfer function of a target filter. The fractional-band compressor section 502 has a side signal path which includes weighting with a target filter 504 with a transfer function H_MB__TAR(Z) and n signal analyzing paths connected to and downstream of the target filter 504. The signal analyzing paths each include a band filter 505 with one of transfer functions H_BF__I(Z) to H_BFJ(Z), a band compressor block 506 with one of compressor functions BLIMi to BLIM_n, and a parameter map block 507 with one of parameter maps BPAR MAPi to BPAR MAP_n. The fractional-band compressor section 502 further includes a main signal path which includes an optional delay unit 508 with a delay time MB delay, and n parametric filters 509 with transfer functions H_PAREQ__I(Z) to H_PAREQ_J(Z), which are connected in series and downstream of the delay unit 508 and which are controlled by parameter sets ParSet_l(n) to ParSet_n(n) provided by the n parameter map blocks 507. The fractional-band compressor section 502 outputs an intermediate signal which is supplied to the full-band section 503.

[0056] The subsequent full-band compressor section 503 includes in the side signal path a target filter 510 with a transfer function H_FB__TAR(Z), a subsequent compressor block 511 with a compressor function FBLIM and a subsequent parameter map block 512 with a parameter map FBMAP. A main signal path of the full-band compressor section 503 includes an optional signal delay 513 with a delay time FB delay and a variable gain stage 514 whose gain is controlled by a gain parameter gFe(n) from the parameter map block 512. The elements mentioned above in connection with the compressor module 501 are described below in detail.

[0057] Target filter 504 provides frequency dependent weighting of the audio signal before the audio signal is evaluated within the side bands (frequency bands in the side signal path), which forms the basis for control of the filters in the main signal path. Target filter 504 sets the frequency dependent target curve for the compressor module 501, which defines the level that leads to signal attenuation (or as the case may be signal amplification) within the main signal path. If the complete fractional-band compressor section 502 is set up accordingly, no signal that exceeds the inverse of a curve normalized to a certain signal level will pass the compressor. However, such a set-up may lead to unpleasant results due to extreme tonality modifications resulting from the unrestricted frequency selective attenuation. The target filter 504 may be implemented as finite impulse response (FIR) or infinite impulse response (IIR) filter with an arbitrary number of tabs or biquads. If implemented as FIR filter, it may be designed as linear phase (constant group delay) filter to avoid changes in the phase relation of the signal.

[0058] Band filters 505 split the weighted audio signal into n fractional (side) bands. It should be noted that n may also be 1, which means there is only a single fractional band. The fractional bands may not cover the full audio bandwidth. If the limited frequency response of the fractional-band compressor section 502 is required to be flat (for a flat target curve), the band filters in the side signal path(s) and the corresponding parametric filters in the main signal path should match. Matching includes adjusting the center frequency and, to a large extent, the bandwidth but also the complete shape of the filter curves. To avoid insufficient limitation at the frequencies that fall between the filter bands, it may be required to have slightly higher bandwidth or a filter curve shape that applies less attenuation at the sides of the bands for the band filters than for the filters in the main signal path. Adequately designed FIR filters or multiple IIR biquads may provide the required transfer function. A sufficiently flat limited response can also be achieved with standard IIR band-pass filters in the side signal path(s).

[0059] Compressor blocks 506 are nonlinear gain elements (e.g., gain changes nonlinearly with input signal level) with characteristic curves (compressor function) that may generally have arbitrary shapes. In addition, these elements may include time constants that control the onset and the slope of any gain change, whereby instead of the attenuated audio signal the momentary gain will be output to the next stage. To calculate this gain it may, depending on the implementation, actually be required to apply it to the audio signal inside the compressor block. However, as only the gain is passed on for further processing, it is irrelevant what is done with the audio signal. An example for such a compressor block from which the gain can be taken may be found, for example, in US8494182B2 (e.g., Figures 11 and 12).

[0060] Parameter map blocks 507 convert gain signals gBl(n) to gBn(n) from the corresponding compressor blocks 506 into parameters ParSet_l(n) to ParSet_n(n) which define the transfer function of the parametric filters 509 in the main signal path. Dependent on the implementation of the parametric filters 509, the parameter sets may include all or at least a subset of the IIR or FIR filter coefficients that define the parametric filters 509. The parameter maps may be implemented as simple look-up tables that choose a set of filter coefficients based on the incoming gain. A two-stage approach is also possible with a look-up table or some kind of linear or nonlinear function that maps incoming gain values to other gain values in the first stage. For example, negative gain values of a certain range may be mapped to positive gain values while negative values from another range are mapped to other negative values by nonlinear or piecewise linear functions. In the second stage the filter coefficients are calculated. The function of the second stage may also be carried out by an additional external coefficient calculation block if the gain values are to be evaluated or modified after calculation.

[0061] Parametric filters 509 are used to attenuate or, as the case may be, amplify the associated frequency band. Parametric filters 509 may be IIR and FIR filters. As described above, the parametric filters 509 should match the band filters 505 of the side signal path if the limited frequency response is expected to be flat for a flat target curve HMB_TAR(z). In this case the filters should also be complementary, in the sense that they add up to a flat response curve if configured with the same positive or negative gain. It should be noted that a flat limited frequency response is not mandatory for the application of the compressor and that some applications may explicitly require a non-flat limited frequency response for a flat target curve.

[0062] Delay unit 508 is used to restrict or avoid overshoots of the fractional-band compressor section 502. Delay unit 508 can be used to compensate any delay caused by the signal processing in the side path(s) and, in addition, any time required to ramp the coefficients of the parametric filters in the main signal path. In some cases it may be beneficial to allow a certain overshoot. Then the delay time may be reduced or set to zero. In other cases, however, overshoot in the fractional-band section 502 is not desired as the subsequent full-band section 503 will react to the overshoot signal and apply additional limitation that may not be desired. Therefore, it may be of some benefit to allow overshoot only in the full-band compressor section. The delay may be substituted by an all-pass filter.

[0063] Target filter 510 provides frequency dependent weighting of the intermediate signal before it is evaluated by compressor block 511. Target filter 510 sets the frequency dependent target curve for the full-band compressor 503 which defines the level that leads to signal attenuation (or amplification) in the main signal path. If the full-band section 503 is set up accordingly, no signal that exceeds the inverse of the curve normalized to a certain signal level will pass the compressor. The full-band section 503 may be set up in this way to prevent any excessive signal levels from being applied to the loudspeaker. One exception would be temporally limited overshoots that can be allowed under certain circumstances. The target filter 510 may be implemented as FIR or IIR filter. If implemented as FIR filter it may be designed as linear phase (constant group delay) filter to avoid changes in the phase relation of the signal.

[0064] Compressor block 511 is a nonlinear gain element (e.g., gain changes nonlinearly with input signal level) with a characteristic curve that may have arbitrary shapes. Compressor block 511 may include time constants that control the onset and the slope of any gain change, whereby instead of the attenuated intermediate signal the momentary gain will be output to the next stage. The calculation of this gain may include applying the gain to the intermediate signal within compressor block 511. However, as only the gain is passed on for further processing, it may be irrelevant what is further done with the intermediate signal in the side signal path of full-band section 503. An example for such a compressor block from which the gain can be taken may be found in US8494182B2 (e.g., Figures 11 and 12).

[0065] Parameter map block 512 maps gain signal gF(n) from the preceding compressor block 511 to a gain value gFB(n) applied to the variable gain element 514 in the main signal path. It is usually implemented as a look-up table but can also be implemented as some kind of linear or nonlinear function. For example, negative gain values of a certain range may be mapped to positive gain values while negative values from another range are mapped to other negative values by nonlinear or piecewise linear functions. The variable gain stage 514 reacts instantly to the gain value on its control input, which is gain value gFB(n). [0066] Optional delay unit 513 may be used to avoid overshoots of the full-band compressor section. It may therefore be used to compensate any delay caused by the side signal path processing as well as any time required to ramp the gain of the variable gain stage. The delay unit 513 may be substituted by an all-pass filter.

[0067] As noted previously, the combination of the fractional-band compressor section 502 with the full-band compressor section 503, which results in the complete multi-band compressor module 501, holds advantages that cannot be achieved by the separate stages. For example, if the fractional-band section 502 alone is used to fully avoid excessive signal levels that could lead to annoying distortion or even destruction of the loudspeaker to be protected, the tonality of the audio signal will be altered significantly. The reason for this is that every frequency band is treated separately from the other bands. Therefore, the bands that cause the highest excursion or power consumption of the loudspeaker will be limited much more strongly than frequency sections with low energy. In practice this means that in many musical compositions the low frequency region (bass) will be suppressed almost completely while the higher frequency bands remain unaltered. The result is highly unpleasant sound.

[0068] Another problem arises from the fact that the signals in the separate bands will ultimately add up to give the full signal level supplied to the loudspeaker. But the total signal level is not available in any fractional-band. Therefore, the resulting total level may still be too high even if all fractional-bands are within their allowable range.

[0069] If the full-band compressor section 503 alone is used to fully avoid excessive signal levels that could lead to annoying distortion or even destruction of the loudspeaker to be protected, it will apply its gain reduction to the whole frequency band. The required gain reduction is determined by the frequency bands that cause the highest excursion or power consumption of the loudspeaker. These frequency bands will be in the lower range of the audible spectrum, usually forming the bass notes of a musical composition. This means that the volume of the mid to high frequency content of the musical composition will be altered synchronously with the bass notes. This annoying phenomenon is typical for full-band compressors and is known as pumping or breathing.

[0070] A further drawback of full-band compressors arises from the fact that the time constants, which define the time after which the compressor starts to change the gain as well as the slope of the gain change, are equal over the whole frequency band. Time constants which may have positive effects for low frequencies, as they allow a slight overshoot which may lead to better perceived audio performance, can easily be too long for higher frequencies. A compromise between these concurring requirements will exhibit some drawbacks.

[0071] Finally, full-band compressors reduce the level of the full frequency band equally while changes at the lower and upper ends of the audible frequency spectrum are less audible than in the mid bands. Thus, the audibility of the compressing effect is higher than for multi- band compressors.

[0072] The combination of the fractional-band section 502 with the full-band section 503 allows alleviating the respective drawbacks caused by each of these sections alone. In the first stage, fractional-band compressing will apply stronger compression to the lower and higher ends of the frequency spectrum with time constants optimized for the respective frequency range, thereby reducing the audibility of the compression. In addition, perceived total sound pressure level during music playback will often be higher than for a full-band compressed signal as the music content within the frequency range that the human ear is most sensitive to will receive less attenuation than would be the case for a full-band compressor. Frequency bands that cause excessive excursion or power loss in the loudspeaker are suppressed more strongly than bands with low energy. However, as the tonality may not be altered too much, the compression that the fractional compressor applies has to be restricted and thus it cannot cover the total required gain reduction. In the second stage, full-band compression will further reduce the gain until a safe level for the loudspeaker to be protected is reached. Nevertheless, as this limitation is considerably lower than it would be without the fractional-band compressor, the negative effects of full-band compressing are reduced significantly.

[0073] The multi-band delay 508 in the fractional-band section 502 of the compressor module 501 with serial signal flow structure effectively separates the fractional-band section 502 chronologically from the full-band section 503, thereby allowing the fractional-band section 502 to compress the signal before the full-band section 503 analyzes it, i.e., the compression of one of any two chronologically separated sections is applied at least partly before the other section analyses the signal. The same applies for the full -band delay 611 in multi-band compressor modules 601 with parallel signal flow structure as set forth further below. [0074] Another option for partly separating the fractional-band section 502 from the full- band section 503 is the use of different thresholds in both sections. In this way the fractional- band section 502 can, for example, compress signals in the low and high frequency bands, as determined from a human audio perception model (also referred to as psycho-acoustic perception model) such as equal loudness curves, to well below the limits of the sound pressure level (SPL) capabilities of the loudspeaker. As a result, the tonality is preserved when the volume is increased. In this case the full-band section 503 is configured to apply attenuation only if required by the loudspeaker and for a certain SPL range only the fractional-band section 502 will provide attenuation.

[0075] If an identical target curve is used for the fractional-band section and the full-band section, the signal flow may alternatively employ a parallel signal flow structure as shown in Figure 6. This structure is particularly useful when a small number of fractional bands is used because the processing required for the target curve may be considerably higher than for parametric filters. This advantage, however, comes with the drawback of increased delay time in the multi-band delay, which needs to be as high as the sum of the delays of the multi- band delay and the full-band delay in a serial signal flow setup. While this does not alter the total delay applied to the music signal it requires more memory space as the full-band delay is still required.

[0076] In an exemplary multi-band compressor module 601 shown in Figure 6, digital input audio data x(n) are input to the compressor 601 and the processed audio data are output as digital output audio data y(n). Compressor 601 combines a fractional-band section 602 with a full-band section 603. The fractional-band section 602 has a side signal path which includes a target filter 604 with a transfer function H_MB__TAR(Z) and n signal analyzing paths connected to and downstream of the target filter 604. The signal analyzing paths each include a band filter 605 with one of transfer functions H_BF__I (Z) to H_BFJ(Z), a band compressor block 606 with one of compressor functions BLIMi to BLIM_n, and a parameter map block 607 with one of parameter maps BPAR MAPi to BPAR MAP_n. The fractional-band section 602 further includes a main signal path. The main signal path includes an optional delay unit 608 with a delay time MB delay and n parametric filters 609 with transfer functions H_PAREQ__I (Z) to HpAREQ_n(z), which are connected in series and downstream of the delay unit 608 and which are controlled by parameter sets ParSet_l(n) to ParSet_n(n) provided by the n parameter map blocks 607. [0077] The full-band section 603 includes in a side signal path a full-band delay 611, n series-connected parametric filters 610 with transfer functions H'_PAREQ__I(Z) to H'_PAREQ_J(Z), a compressor block 612 with the compressor function FBLIM and a subsequent parameter map block 613 with the parameter maps FBMAP. A main signal path of the full-band compressor section 603 includes a variable gain stage 614 which is controlled by the gain parameter gFB(n) from parameter map block 613. In the side signal path, the n parametric filters 610 are connected in series and downstream of the target filter 604 and are controlled by parameter sets ParSet_l(n) to ParSet_n(n), provided by the n parameter map blocks 607. The transfer functions H'_PAREQ__I(Z) to H'_PAREQ__II(Z) of parametric filters 610 may be identical with transfer functions H_PAREQ__I(Z) to H_PAREQ_J(Z) of parametric filters 609.

[0078] The elements mentioned in connection with the compressor 601 perform identical or similar functions as those described above in connection with the corresponding blocks of multi-band compressor module 501 shown in Figure 5.

[0079] The lowest complexity is reached if the number of fractional-bands is reduced to N = 1. Figure 7 shows this option with a parallel signal flow although a serial signal flow would be possible as well. It is still valid to designate such a compressor in total as a multi-band compressor because it processes one fractional -band and the full-band. While multiple fractional-bands yield improved results, it is nevertheless possible to design a single fractional-band section in a way that is of some advantage over a simple full-band compressor.

[0080] As mentioned before, the compression applied to the fractional-bands should be restricted in order to keep tonality changes within certain tolerances. For example, tolerable tonality modifications can be derived from curves of equal loudness perception as shown in Figure 8 which depicts the equal loudness curves from ISO 226:2003. The loudness curves can be referenced to a curve of a certain level as shown in Figure 9. It should be noted that the curves for 110 and 120 phon have been extrapolated from the lower curves as these signal levels are not available from ISO specifications. In the present example, 80phon was chosen as reference level because this is a typical listening level for which music content may be optimized. Different reference levels may be chosen for other applications. It can be seen that for a perceived loudness increase that equals an increase of lOdB above the 80 phon level at 1kHz, the SPL only needs to be increased by 4.7dB at 20Hz. This means that for a broadband volume increase of lOdB from 80dB SPL at 1kHz to 90dB SPL at 1kHz the gain at 20Hz may be decreased by 5.3dB without inducing a tonality change between the two volume levels. The equal loudness curves show how much compression is possible over frequency without change of overall tonality. Therefore, the slope of the curves and the compression compared to 1kHz can be used to derive restrictions for tonality changes induced by frequency selective compression.

[0081] As shown in Figure 10, the curves above 80 phon may be shifted downwards by their nominal SPL increase while curves below 80 phon may be removed (curves are named by the gain shift applied). Intermediate curves have been interpolated with 2dB step size. The resulting curves show how much attenuation per frequency can be applied as a reaction to a certain total volume increase without causing tonality changes. For example, if the volume is increased by 20dB, the acceptable frequency selective compression at 20Hz is -10.6dB. For a 30dB volume increase the acceptable frequency selective compression at 200Hz is -6.1dB.

[0082] For any frequency, an equal loudness factor kEoj f) can be determined from the curves shown in Figure 10 that approximates the ratio between required compression and compression without tonality change with sufficient accuracy. For example, the total attenuation range of the curves at 20Hz is 15.9dB for a total range of required attenuation of 30dB, resulting in an equal loudness factor of ICEQL (20HZ) ~ 0.53. This factor may be applied to the required attenuation which will return the allowable attenuation for the fractional-band compressor. It should be noted that the ratio of perceived loudness change versus SPL change as described by the equal loudness curves is approximately linear over a certain SPL range but tends to become increasingly nonlinear towards the extremes of the dynamic range of the human auditory system. If the SPL range of the application includes a nonlinear region of the human auditory system, it may be beneficial to include the nonlinearity in the equal loudness factor. In this case ICEQL may be defined as nonlinear function of the required attenuation gTOT(f) in the form of kEQi f, gTOT(f))). As an example based on the curves of figure 10 this means that k_EQL (_20Hz, - dB) * = 0.525 for g_To_T(20Hz) = -4dB and k_EQL (20Hz, -26dB) * _ ^ * 0.531 for g_To_T(20Hz) = -26dB. As the ratio of perceived loudness change versus SPL change is approximately linear for the dynamic range of figure 10, kEQL is almost equal for both values of gTOT. For the following descriptions and formulas kEQL(f) may be replaced by kEoj f, gTOT(f)). [0083] As the multi-band compressor module combines fractional-band compressing with full-band compressing, the difference between the attenuation applied by the fractional-band section and the required attenuation will additionally be attenuated in the full-band section if the target curves and compressor block thresholds are the same in the fractional-band section and the full-band section. It should be mentioned as an exception that if no compression is applied in the full band section, the result is dynamic loudness compensation above the given threshold for compression in the fractional-band section. If the full-band section limits the signal this will cause tonality changes. How much tonality change is allowable for any given total required attenuation gTOT(f) is determined by the frequency selective equal loudness factor kEQL(f), wherein the applied attenuation gAPp(f) is calculated as:

[0084] g_APP {f) = k_EQL {f) * g_T0T (f)

[0085] It should be noted that there is a tradeoff between perceived total loudness and tonality. The more attenuation is applied to low frequencies the less attenuation is required in the midrange, leading to higher total loudness. Tonality, however, will also change to an increasing extent as more attenuation is applied in the low frequencies. Dependent on the application, either higher total loudness or lower tonality modification will be preferable. The limitation restriction may be altered to various extents to best suit the application. For this purpose a factor kroN can be established that controls the extent of tonality change:

[0086] g_APp (f) = Jk_T0N * k_EQL(f) * g_T0T(f),

wherein kroN can also be made frequency selective as kTON(f) to allow, for example, higher attenuation at frequencies below the protected loudspeakers resonance frequency to avoid membrane excursions that, in case of bass reflex or passive radiator systems, do not generate sound.

[0087] As noted before, kEQL can be a linear or nonlinear function of gTOT, for example, in the form of kEQL(f,gTOT(f,n)), which is recalculated for any audio sample n and integrated in the calculation of gAPp(f,n) as:

[0088] g_APp (f, n) = ^jk_T0N * k_EQL (f, g_T0T f, n)) * g_T0T (f_> n) .

[0089] With kEQL(f,gTOT(f,n)) it is possible to set an absolute restriction for tonality changes. The factor may, for example, be kept constant for values of gTOT(f,n) above a certain threshold value and increase at the same rate at which gTOT(f,n) decreases to below the threshold. In addition kEQi f, gTOT(f,n))) may also include any nonlinearity of the equal loudness factor concerning gTOT(f,n) as described above for the extremes of the dynamic range of the human auditory system. An exemplary implementation may employ a look-up table which maps the required total attenuation gTOT(f,n) to the applied attenuation gAPp(f,n). Alternatively or additionally, other psychoacoustic models may be chosen do determine the restrictions of fractional-band compression. It is, for example, possible to allow higher attenuation for narrow frequency ranges than for wider ranges. By means of the masking effect, tones around narrow frequency bands will effectively lower the audibility of narrow band attenuation. This can be used to suppress frequencies that are especially prone to certain sound distortions or artifacts (e.g., rub and buzz). It is also possible to determine the restrictions that empirically lead to the best results for any given application.

[0090] As the target curves implemented in the target filter define the maximum voltage levels that will pass the compressor in a steady state (in some cases overshoots may be allowed during the attack phase of the compressor), these curves are derived from the loudspeaker that is to be protected. Several options for determining the target curve can be chosen, such as but not limited to an empirical approach with listening tests, automated acoustic measurements with evaluation based on psychoacoustic models, total harmonic distortion and noise measurements, membrane excursion measurements, power measurements, SPL measurements of the loudspeaker, excursion and power modeling of the loudspeaker.

[0091] In an empirical approach, listening tests may be conducted to determine which signal level leads to annoying distortion at any frequency of interest. For example, sine waves or more practically sine bursts may be used as test signals, and also multi-sine test tones covering a certain frequency range may be employed. The use of sine waves allows an easy identification of distortions. In this context, distortion refers to any audible sound other than the test signal. For example, the rubbing of the voice coil in the air gap, caused by a rocking motion of the voice coil and the membrane assembly, may produce a buzzing noise well before the maximum excursion is reached. Distortions may also arise from port noise generated inside bass reflex ports at high air velocities and from parts outside the loudspeaker which vibrate due to excitation at certain frequencies, thereby emitting noises, as for example elements of the loudspeaker cabinet or parts of car interiors. As distortion increases when the excursion of the voice coil and the membrane approaches the maximum excursion of the loudspeaker, the excursion is inherently limited. [0092] Automated acoustic measurements with an evaluation based on psychoacoustic models are conducted at any frequency of interest and evaluation of distortion components of any kind may be used. For example, sine burst signals of increasing levels may be applied to the loudspeaker under test, the resulting acoustic signals being recorded with a microphone. For evaluation of the distortion, the test frequency may be moved into the time domain, e.g., by way of steep filters, or into the spectral domain e.g., by way of a Fast Fourier Transformation (FFT) with subsequent deletion of the respective spectral component. The remaining signal represents the distortion generated by the loudspeaker. Psychoacoustic models like the equal loudness curves and the masking effect may be used to shape the distortion signal and to determine the annoyance level.

[0093] For total harmonic distortion and noise measurements, the maximum voltage level, at which the total harmonic distortion (THD) or the total harmonic distortion with noise (THD+N) is below a certain limit, is measured at any frequency of interest. The limit may itself vary over frequency, e.g., in a way that increases the allowable distortion towards lower frequencies. For intermodulation distortion measurements, the maximum voltage level, at which the intermodulation distortion IMD caused in combination with one or multiple tones at higher frequencies stays below a certain limit, is measured for any frequency of interest. The limit may itself vary over frequency, e.g., in a way that increases the allowable distortion towards lower frequencies.

[0094] For membrane excursion measurements, automated or manual measurements at any frequency of interest are conducted to measure the maximum voltage applicable to the loudspeaker within its safe excursion range. The measurements may be conducted through laser tracking or with acceleration sensors attached to the loudspeaker membrane.

[0095] For power measurements, automated or manual measurements of the power applied to a loudspeaker are conducted at any frequency of interest to determine the maximum voltage level that causes the maximum allowable power loss in the loudspeaker voice coil.

[0096] For measurements of the loudspeaker's SPL it may be desirable to limit the sound pressure level to a certain threshold value below the loudspeaker's allowable maximum SPL. Measurements of the loudspeaker's SPL over frequency can be used for target curve generation if the sound pressure level is limited to a certain threshold value below the loudspeaker's allowable maximum SPL. This may be useful for headphones to avoid damage of listeners' ears. The SPL vs. driving voltage curve may be measured and based thereon a target curve may be set to avoid excessive SPL. In addition, the magnitude response (SPL vs. frequency) may be used to generate the target curve for loudness compensation. If excursion and power modeling of the loudspeaker are employed, the modeled excursion and power loss of the loudspeaker may be measured for any given frequency and voltage level to derive the allowable voltage per frequency and its corresponding target curve.

[0097] It may be beneficial to use a combination of the methods described above, either for the full frequency band of interest or for parts of it, to define the target curve. In addition, amplifier clipping may be avoided in that the highest possible voltage is set below the amplifier's clipping voltage by appropriate definition of the target curve. If the amplifier is included during the determination of the target curve by empirical approach with listening tests, automated acoustic measurements with evaluation based on psychoacoustic models or total harmonic distortion and noise measurements, clipping will only occur to an extent that does not violate the tolerance threshold of the test. In case of automated testing, a larger number of samples may be evaluated to capture production tolerances. Concerning the resonance frequency of a dynamic loudspeaker disposed inside an enclosure, safety margins should be applied, as the resonance frequency may shift over temperature and due to aging and production tolerances.

[0098] Based on the structure shown in Figure 4 in connection with a series signal flow as shown in Figure 5 and with four band filters in the fractional-band section, exemplary settings of some compressor elements are explained below. It is assumed that both target filters 504, 510 have the same transfer function so that HMB_TAR(Z) = HFB_TAR(Z).

[0099] The implemented target curve may be determined by the empirical approach based on listening tests. During testing, both loudspeakers 301, 302 may be driven by the same sine burst signals and the level may be increased until distortions are considered to be annoying. The resulting target curve may be approximated with a 10 biquad infinite impulse response (IIR) filter structure, for which the exemplary amplitude vs. frequency plot is shown in Figure 11. As the highest peaks in the amplitude response will lead to the strongest limitation of signal level, the curve shown in Figure 11 represents the inverse of the tolerable voltage curve.

[00100] All parameter maps may be implemented as look-up tables with subsequent IIR filter parameter calculation. The restriction of the limitation applied in the fractional-band section 502 may be controlled by the combination of the filter type and parameters chosen for the parametric filters 509 in the main signal path and the gain mapping performed by the look-up tables in the parameter map blocks 507. Gain mapping may employ interpolation for gain values not listed in the look-up tables. Filter type, frequency and quality may be fixed and filter gain may be updated based on the gain values from the preceding compressor blocks 506, which have been converted to the filter gain by the look-up tables. In Figure 13, the resulting attenuation over frequency for singular stationary sine tones is shown. The exemplary graph relates to a flat target curve and compressor block settings that lead to - 20dB calculated compressor gain in the compressor blocks.

[00101] For the band filters that separate the fractional-bands, the frequency response shown in Figure 12 may be implemented by way of IIR filters. All compressor blocks may be implemented based on a similar structure as described in US8494182B2. Delay, attack and release times are chosen appropriately for the frequency range of the corresponding band. All compressors are set to the same threshold value, which is chosen to deliver the required limited voltage levels at the loudspeaker with the given overall multi-band compressor module settings and the power amplifier. From the graph shown in Figure 13, it can be seen that the slope of the low frequency limitation (<200Hz) resembles the restriction derived from the equal loudness curves for -20dB required overall limitation as depicted in Figure 10.

[00102] In this example, all parametric filters, e.g., parametric filters 509 with transfer functions HPAREQ_I(Z) to HPAREQ_₄(Z), shown in Figure 5, are implemented in a single biquad IIR filter structure. The filter coefficients are supplied by way of parameter maps as described above. The delay unit 508 in the fractional section 502 may be chosen to avoid overshoots in the fractional-band section 502, which may lead to excessive compression or limitation in the full-band section 503. The parameter map 512 for the full-band section 503 is omitted (bypassed) in this example. All gain values for the variable gain stage 514, which are calculated by compressor block 511, may be directly applied by the gain stage 514. The delay unit 513 delay in the full-band section 503 may be chosen empirically to allow some overshoot that leads to more dynamic sound without causing excessive excursion or distortion.

[00103] In order to achieve a finer granularity of the fractional-band compressor section, it is desirable to use a higher number of frequency bands, especially at the low to mid frequency range. In an example, a total number of 12 parametric filters may be controlled by parameter maps to track the compression or limitation restrictions determined by the equal loudness curves. Figure 14 shows the corresponding amplitude frequency response of all parametric filters as well as their sum (parametric filters) and the target slope determined from the equal loudness curves for a total limitation of -30dB. Tracking of the filter slope defined by the equal loudness curves is advantageous, as can be seen in Figure 15 which shows matching for multiple limitation levels.

[00104] Tracking of the attenuation restrictions may be implemented by use of gain factors for every fractional-band. One set of gain values may be used to correct the filter bank, which in the present exemplary case may include 11 equalizing (EQ) filters and one second order high- shelve filter, to give a largely flat response over the audible range for uniform attenuation factors on all bands. Another set of gain factors (EQL mapping) represents the attenuation applied per band for a given total attenuation requirement from the preceding compressor block.

[00105] In another example shown in Figure 16, multiple compressor blocks BLIM are connected in parallel so that multiple thresholds and compressor characteristics (e.g. compression ratio or characteristic curve, attack time, release time, hold time, etc.) can be implemented per frequency band. The multi-band compressor module 1601 shown in Figure 16 is based on the multi-band compressor module 501 described above in connection with Figure 5, however, each of the compressor blocks 506 is substituted by (at least) two parallel connected compressor blocks 1602 which have compressor functions BLIMi i, BLIM 1 2, . . . BLIMn 1 , BLIM_n_2. Parameter map blocks 1603 with parameter maps BPAR MAPi 1, BPAR MAP 1 2, BPAR MAP_n_i, BPAR MAP 1 2 are connected downstream of the at least two parallel connected compressor blocks 1602 and substitute the parameter map blocks 507 in the multi-band compressor module 501 shown in Figure 5. The outputs of two corresponding parameter map blocks 1603 are combined by a coefficient calculation block 1604 to provide parameters ParSet_l(n) to ParSet_n(n). Furthermore, each of the compressor blocks 511 is substituted by (at least) two parallel connected compressor blocks 1605 which have compressor functions FBLIMi 1 and FBLIMi 2. Parameter map blocks 1606 with parameter maps FBMAPi 1 and FBMAPi 2 are connected downstream of the at least two parallel connected compressor blocks 1605 and substitute the parameter map blocks 512 in the multi- band compressor module 501 shown in Figure 5. The outputs of two corresponding parameter map blocks 1606 are combined by a coefficient calculation block 1607 to provide gain parameter gFB(n). In this case, the parameter maps BPAR MAP and FB MAP only implement gain mapping while the combination of the resulting gains (e.g., minimum gain) and the calculation of the corresponding filter coefficients is accomplished in the subsequent coefficient calculation block.

[00106] Multiple target curves may be required if the multi-band compressor module is to be used to control different parameters separately. Different target curves may be used for voltage, current, power and thermal limitation or for loudness compensation. For the latter, the target curve may be set to follow the magnitude response of the attached loudspeaker. In this way the SPL that results from signal processing upstream of the multi-band compressor module and the loudspeaker response is used for loudness compensation. In the multi-band compressor module 1601 shown in Figure 16, target filter 504 may be substituted by (at least) two target filters 1704 with transfer functions H_MB__TAR__I(Z) and H_FB__TARJ(Z) to form a multi- band compressor module 1701 with a fractional-band section 1702 and a full-band section 1703 as shown in Figure 17. Furthermore, target filter 510 may be substituted by (at least) two target filters 1705 with transfer functions H_FB__TAR__I (Z) and H_FB__TAR_₂(Z).

[00107] Referring to Figure 18, in a multi-band compressor module 1801 with a fractional- band section 1802 and a full-band section 1803, which is based on the a multi-band compressor module 501 shown in Figure 5, at least one filter 1804 of the parametric filters 509 in the main signal path of the fractional-band section 1802 may be connected in parallel (e.g., employing an adder 1805) for various functions. For example, bass boost and low frequency limitation may be implemented with different filters. The control of the filter 1804 may use a separate side band 1806 with respective target filter, compressor block and parameter map block.

[00108] Alternatively, a combination of multiple sidebands may be used to control the filter 1802 as in the multi-band compressor module 1901 with a fractional-band section 1902 and a full-band section 1903 as shown in Figure 19. To combine multiple sidebands, a combiner 1904 and a subsequent parameter map block 1903 may be employed.

[00109] A multi-band compressor structure allows boosting of arbitrary frequencies without risking the destruction of the loudspeaker, so that dynamic bass boost or dynamic loudness, in form of frequency selective boost for playback levels below a certain reference level, can be applied to the loudspeaker signal. Limiter blocks may evaluate whether and to what extent the loudspeaker signal exceeds a given threshold and may calculate a gain that prevents the threshold from being exceeded. The threshold is, for example, set to a level that marks the level above which limitation is required for loudspeaker protection. Dynamic changes of the limiter gain may be controlled by different time constants for increasing and decreasing gain. If the compressor block is configured to have a threshold value much lower than any threshold required to protect the loudspeaker, it will permanently calculate changing gain values that decrease quickly if the audio signal increases and increase slowly if the audio signal decreases. By way of mapping, it is possible to use a certain range of gain values for actual gain reduction of the parametric filters in the main signal path of the fractional-band section while another range of values can be used to increase the gain of these filters. As a result, the frequency range controlled by the respective filters is boosted if the audio signal is below a certain threshold for a defined period of time and is attenuated if it is above that threshold. This threshold may be considered as the reference level for loudness compensation or bass boost. A look-up table may be used for bidirectional calculation (attenuation and amplification) of the gain for the parametric filters (filter gain) based on different ranges of the gain values from the compressor block. Mapping of intermediate gain values may include some kind of interpolation.

[00110] As is the case for tonality changes induced by frequency selective limitation, bass boost should not be applied excessively. Otherwise the overall tonal balance may be lost and the bass notes will be exaggerated. For example, the tolerable tonality modifications can be derived from curves of equal loudness perception in the same way as described in connection with Figures 8, 9 and 10. The result is shown in Figure 20 which depicts the allowable amplification per frequency for a given headroom between actual audio signal level and loudspeaker voltage handling abilities or, generally, the reference level for loudness compensation or bass boost (one curve per 2dB headroom increment). As was the case for tonality changes, the amount of boost applied per frequency can be adjusted during limitation. If the amount of applied boosting per frequency is equal to the difference between equal loudness curves the result is loudness compensation. Loudness compensation can be done for specific frequency ranges or over the whole audible frequency band. It should be noted that, for better control of loudness compensation results, the target curve used for loudness compensation may reflect the ratio of signal level to acoustical sound pressure level of the attached loudspeaker (magnitude response). This will essentially result in a reference playback level for loudness compensation that is flat over frequency. The reference playback level over frequency may additionally be shaped by superposition of further magnitude responses. For example, the optionally inverted shape of the equal loudness curve of a certain loudness level (e.g., 80 phon) may be superimposed on the loudspeaker magnitude response. Another option would be to superimpose the inverted target frequency response of the loudspeaker. In this way, an emphasized bass range may also be reflected in the way loudness is compensated, as the respective range will exhibit a higher threshold level for loudness compensation, leading to boosted signal levels over a wider range of playback levels. Generally, any magnitude response may be superimposed on the target curve to adjust the reference level for loudness compensation and thereby the acoustic result. Good results can also be obtained with a single target curve for loudness compensation and protection that represents the protection threshold of the loudspeaker over frequency.

[00111] Especially in the bass range, it may be desirable to add more boost than would be required for loudness compensation. This can be achieved in the same way as described in connection with Figures 7-9 by use of a frequency selective factor such as factor kroN(f). In addition, different filter characteristics for the parametric filters in the main signal path may be used for limitation and boosting. For example, it may be desirable to limit the bass boost applied to a certain frequency range in order to avoid boost of frequencies below the lower cutoff frequency of the loudspeaker. Therefore, equalizing (EQ) filters might be better suited for application of bass boost than shelving filters, which may in turn be better suited for limitation. Bass boost may be applied by way of bass extension, whereby essentially the frequency response of the loudspeaker at lower frequencies is extended, retaining the basic magnitude response characteristic of the loudspeaker (e.g., flat magnitude curve over frequency) if the dynamic capabilities of the loudspeaker allow for it. This may be done with a peak EQ filter that boosts the relevant frequency range around the lower cutoff frequency of the loudspeaker. In this case, the boost applied for loudness compensation and bass extension may not exceed the dynamic headroom of the loudspeaker. This may be ensured with look-up tables that control bass extension and loudness compensation filters in the way that the dynamic headroom of the loudspeaker is essentially shared between both functions. If the peak EQ filter used for bass extension is also controlled to attenuate the respective frequency range in case the dynamic capabilities of the loudspeaker are exceeded by the audio signal, the low frequency extension of the loudspeaker is reduced. Low frequency range extension and reduction can be performed in parallel with loudness compensation without exaggeration of bass notes. [00112] When the compressor protects the loudspeaker by way of frequency selective attenuation of the audio signal, it may be desirable for low frequencies to replace the attenuated spectrum of the audio signal by virtual bass. In this context, virtual bass means the augmentation or total replacement of the original bass signals with higher frequency harmonics that can be played by the loudspeaker due to lower required excursion. For example, a phase vocoder may be used to generate the harmonics. In this case, weighting of the harmonics can be done based on the attenuation that the compressor applies to the audio signal. Figure 21 shows exemplary connections between the harmonics generator, e.g., vocoder 2101 and the multi-band-compressor module 501 as shown in Figure 5. The harmonics generator receives the unaltered audio signal and the information on how much attenuation is required per frequency band, as well as the filter coefficients after gain mapping. Alternatively, a part of the gain mapping inside the multi-band compressor module may be replicated in the harmonics generator to obtain the actual attenuation per frequency band. Weighting of the harmonics is implemented in the harmonics generator in the way that harmonics are amplified by a factor derived from the attenuation of the underlying fundamental frequency. The generated harmonics are then added to the audio signal that is to be input into the full-band compressor section. Alternatively, the harmonics may be added upstream of the fractional-band compressor section.

[00113] Another way to control which frequency bands are primarily used for harmonics generation is described below in connection with Figure 22. Inverse filtering of the audio signal x(n), in contrast to the filtering applied in the multi-band compressor module 501, is used to determine which frequency bands are supplied to an harmonics generator 2202. The harmonics generator 2202 outputs a signal that is added to the audio signal fed into full-band section 503 and may be based on a phase vocoder, a nonlinear device or a combination thereof. The required attenuation per frequency from any or all of the compressor blocks 506 with compressor functions BLIMi to BLIM_n-i of the fractional-band section 502 is supplied to parameter map blocks 2203 ( parameter maps BPAR MAPi' to BPAR MAP_n-i') in a virtual bass module 2201. Which frequency bands are included in the virtual bass generation depends on their distribution on the frequency axis and the type of loudspeaker to be controlled. For example, low frequency bands (<300Hz) may be used for virtual bass generation. [00114] The parameter map blocks 2203 in the virtual bass module 2201 calculate coefficient sets which correspond to transfer functions H_PAREQ__I'(Z) to H_PAREQ_J-_I'(Z) of parametric filters 2204. The transfer functions H_PAREQ__I'(Z) to H_PAREQ_J-_I'(Z) are the inverse of the corresponding transfer functions H_PAREQ__I (Z) to H_PAREQ_J-_I (Z) in the multi-band compressor module 502. Alternatively, the amplification of the virtual bass module filters may be scaled by a certain factor to achieve higher or lower amplification levels. If, for example, a particular filter in the fractional-band compressor module 502 is an EQ filter with a center frequency fl, quality Ql and attenuation g (g=-x[dB]), the corresponding inverted filter will have the same parameters, except for the gain of +x[dB] or +x[dB]-ks with ks being a scaling factor. Scaling factor ks may be changed for different values of gain g, e.g., according to a function ks(n) =f(g(n)).

[00115] The resulting coefficients are supplied to the parametric filters 2204 which are disposed in the single signal path of the virtual bass module 2201. A low pass filter 2205 may be connected upstream of the harmonics generator 2202 for initial band limitation. For example, the low pass filter 2205 may be disposed upstream of a gain element 2206 with a gain HarGain. Gain element 2206 may be connected upstream of the harmonics generator 2202, e.g., between low pass filter 2205 and parametric filters 2204, and may be used to adjust the level of harmonics added (e.g. by an adder 2207) to the audio signal that is fed into the full-band section 503. It may be set to a value that produces only inaudible harmonics for flat filter responses of the transfer functions H_PAREQJ'(Z) to H_PAREQ_JI-_I'(Z). When the multi- band compressor module 502 attenuates frequency bands, these bands are amplified by the gain HarGain and subsequently lead to higher levels of harmonics added to the audio signal input into full-band section 503.

[00116] In multi-channel audio systems, it may be desirable to control the way individual channels are limited to avoid excessive changes of the system's tonality or spatial representation. Multiple multi-band compressor modules may be linked to form a multichannel compressor providing coherent limitation for multi-channel audio systems. Multichannel systems may, for example, be active multi-way loudspeakers, a stereo pair of active or passive loudspeakers, surround loudspeaker systems made up of active or passive loudspeakers or any combination thereof. The multi-channel compressor may also be applied to multi-channel loudspeaker systems in automobiles. [00117] Figure 23 is a block diagram of an exemplary multi-channel compressor system 2301 with n (e.g., n=3) multi-band compressor modules 2302-2304, also referred to herein as compressor channels, which send their locally calculated frequency selective and full-band gain to a multi-channel link module 2305. The multi-channel link module 2305 calculates revised gain factors from all gain factors gathered from a plurality of compressor channels and returns the gain factors that are now linked to the multi-band compressor modules 2302- 2304 where these gains are applied. In the example shown in Figure 23, multi-band compressor module 2302 may be connected to an optional digital-analog converter 2306, an amplifier 2307 and a loudspeaker 2308 in the manner described above in connection with Figure 1. Multi-band compressor module 2303 may be connected to an optional digital- analog converter 2309, identical amplifiers 2310, 2311 and identical loudspeakers 2312, 2313 in the manner described above in connection with Figure 4. Multi-band compressor module 2304 may be connected to an optional digital-analog converter 2314, an amplifier 2315, a passive cross-over network 2316, and identical loudspeakers 2317, 2318 in the manner described above in connection with Figure 3. For multi-channel linking via gain factors, the multi-band compressor modules may be adapted in a way that the parameter maps BPAR MAPi to BPAR MAP_n-i only calculate the required gain factor per frequency. The coefficient calculation for the parametric filters may be carried out on linked gain factors received from the multi-channel link module 2305. For this purpose, the multi-band compressor modules may have a multi-module structure 2401 as shown in Figure 24, which may be based on the multichannel module 501 shown in Figure 5 and which may have a fractional-band section 2402 (corresponding to fractional-band section 502) and a full-band-section 2403 (corresponding to full-band section 503).

[00118] In the multi-module structure 2401 shown in Figure 24, local gain factors g_xj_oc(f, n) are provided as outputs (from the parameter map blocks 507) and linked gain factors g_x u_nk(f, n) are used to calculate the filter coefficients for the parametric filters 509 in the main signal path (e.g., by way of coefficient calculation blocks 2404). Similarly, a local full-band gain gxjoc_fb(n) forms an output signal (from parameter map block 512) and a full-band gain gxjink_fb(n) is used to control variable gain block 514. All multi-band compressor modules that are combined by a multi-channel link module should have parametric filters with equal transfer functions in corresponding frequency bands when calculated with the same gain factor. This ensures that the limited magnitude and phase response is equal for the same gain factor per frequency band on all compressor channels. But it is not required that all compressor channels implement the same number of frequency bands. For example, a sub- woofer does not require fractional-bands at higher frequencies. If fractional-bands are not needed in certain compressor channels they may simply not be included in the linking process.

[00119] Linking between compressor channels may be based on comparisons between gain factors per frequency band as well as the full-band gain factor. This means that all gain factors of the same frequency band in the linked compressor channels are compared and a new gain factor is calculated for every compressor channel. Dependent on the type of linking, the gain factors per frequency band may be the same for all compressor channels or differ from each other. Different types of linking may be applied and several compressor channels may be combined in groups with a certain type of linking within the group and another type of linking between n=3 link groups as shown in Figure 25. Link groups 2501-2503 may be used to tightly couple the front channels of a 5.1 surround system (e.g., by minimum linking as described below) with each other and the effect channels with each other. A weaker link may be established between those groups (e.g., unilateral linking, so that the front channels affect the back channels but not vice versa). Exemplary types of linking are described below. Any linking is carried out between gain factors of the same frequency band or between the gain factors for the full-band. The linking methods described below may be combined to form further linking types.

[00120] In the case of minimum linking, the lowest gain of all compressor channels is applied to all channels. In the case of fixed factor linking, the lowest gain of all compressor channels is determined and the difference between this value and the individual channel gain is multiplied by a factor before it is additionally applied to the channel gain. In the case of weighted minimum linking, the gain factors per channel are multiplied by a weighting factor before the lowest resulting gain of all compressor channels is applied to all channels. In the case of variable factor linking, the factor as used with fixed factor linking and weighted minimum linking may be a linear or nonlinear function of the actual gain per channel or the difference between the lowest channel gain and the individual channel gain. In the case of time variable linking, linking can be combined with time constants, controlling the onset and slope of gain changes induced by linking. For example, a certain compressor channel requires a certain gain value for a longer period of time before any other channels are linked to it. Or the gain per compressor channel is averaged over a certain period of time and the average gain is used for comparison to other channels. When linking starts, the slope of the gain factor change on the linked channels is controlled by another time constant. In the case of unilateral linking, certain channels may affect other channels although they are not affected by these channels.

[00121] Current or power applied to dynamic loudspeakers may be limited as well with some modifications to the multi-band compressor structure. These modifications are described below for the multi-band compressor module 501 with series structure, as shown in Figure 5, but can also be implemented for the multi-band compressor module 601 with parallel structure as shown in Figure 6, or with any other suitable structures. In order to implement current limiting in a multi-band compressor structure as shown in Figure 5, output voltage scaling (OVS) blocks and, optionally, root-mean- square (RMS) blocks may be added which calculate the quadratic mean over a certain number of samples. An accordingly adapted multi-band compressor module 501 is shown in Figure 26. As can be seen, an OVS block 2601 is connected upstream of the band filters 505 and RMS blocks 2602 are connected between band filters 505 and compressor blocks 506 in the fractional-band section 502 of multi-band compressor module 501. Furthermore, an OVS block 2603 and a subsequent RMS block 2604 are connected between target filter 510 and compressor block 511 in the full-band section 503 of multi-band compressor module 501. OVS serves to determine the voltage that is actually applied to the loudspeaker. In order to scale the (digital) signal to the resulting voltage across the loudspeaker, the OdBFs output voltage of the digital-analog converter and the gain of the power amplifier are taken into account (see also the description below in connection with Figure 32). As OVS may only include multiplication with a factor, it may be combined with an admittance filter 2605 and 2606, which replace the target filters 504 and 510, respectively, in the multiband compressor module 501, have a transfer function HADM(z) and may be connected upstream of the OVS block 2601 and 2603, respectively. The admittance filters 2605 and 2606 serve to determine the momentary current flow in the loudspeaker. Contrary to a target filter of a peak voltage limiter, an admittance filter implements the inverse loudspeaker impedance function (see also the description below in connection with Figure 33). The loudspeaker current representing signals from the RMS blocks 2602 and 2604 are processed in the compressor blocks 506 and 511 and subsequent parameter maps 507 and 512 define the filter functions for the fractional -band main signal path filters or, correspondingly, the gain for the full-band section 503.

[00122] If power is to be limited, a multi-band compressor structure as shown in Figure 27 may be used, which is based on the structure shown in Figure 26. In the multi-band compressor structure shown in Figure 27, voltage scaling is combined with admittance filtering, however, if voltage scaling requires nonlinear or piecewise linear functions, it may be implemented in a separate block. That may be required if, for example, clipping of the amplifier or any other component in the downstream signal path is to be taken into account. In the fractional-band section 502, the current signal i(n) from the admittance filter 2605, which also includes voltage scaling, is band filtered with band filters 505 and subsequently squared and multiplied with the DC resistance R of the voice coil in non-linear processing blocks 2701. Alternatively, the minimum AC impedance within the audio range may be used instead of the DC resistance, as this will include eddy current losses within the voice coil former. Momentary power signals pBi(n) to pBn(n) output by the non-linear processing blocks 2701 may be processed by optional RMS blocks 2602 to calculate the quadratic mean over a certain number of samples. Similarly, in the full-band section 503, a non-linear processing block 2702 is connected between admittance filter 2606 and optional RMS block 2604. The power signals from RMS blocks 2602/2604 are processed in the compressor blocks 506/511 and the subsequent parameter map blocks 507/512 to define the transfer functions of the main signal path filters 509 in the fractional-band section 502 and the gain of the gain block 514 in the full-band section 503.

[00123] Integration of thermal protection for dynamic or electro-dynamic (i.e., with a field coil instead a permanent magnet) loudspeakers is also possible with some modifications to the multi-band-compressor module described above. Excursion limitation, together with limitation of the voice coil temperature, offers complete protection against loudspeaker damage caused by the driving signal. In a multi-band compressor with power compressor function as described above, short term averaging of momentary power applied to the loudspeaker or calculation of the quadratic mean over short term may be used to avoid temporary overloading which may lead to rapid thermal destruction of the voice coil. Heating of the voice coil of dynamic loudspeakers, however, usually has multiple time constants resulting from the thermal capacity of the voice coil itself and from the thermal capacity of other elements of the loudspeaker, such as the membrane, the motor structure (e.g., magnet, pole piece, etc.), the loudspeaker basket, the internal volume of a loudspeaker box and the loudspeaker box cabinet in combination with the thermal resistance between these elements. Some of these time constants can be small (e.g. the voice coil time constant of a tweeter) while others are quite large (e.g. motor assembly). Furthermore, the temperature of the voice coil and the frequency content of the music signal are not directly related if the time constant of the loudspeaker is larger than the rate of frequency content change in the music signal.

[00124] However, it may be feasible to reduce the overall volume to prevent thermal overload while using the multi-band compressor structure, thus providing protection against excessive excursion and/or sound enhancement. Sound enhancement includes distortion reduction, e.g., by way of selective limitation of frequencies that are prone to distortion, dynamic loudness compensation based on momentary frequency content of the signal, and dynamic bass extension based on the momentary voltage headroom of the loudspeaker. For correct operation of the multi-band compressor structure (e.g., a multi-band compressor module 501), the output signal of the multi-band compressor structure may not be altered dynamically by a thermal limiter, as these signal changes are not taken into account in the multi-band structure. Referring now to Figure 28, an exemplary multi-band compressor with thermal protection characteristics may include a multi-band-compressor module 2801 (e.g., for peak- signal processing only), which is supplied with the input signal x(n). The input signal x(n) may have been preprocessed by variable gain amplifier 2802, which is connected upstream of the multi-band-compressor module and which provides the output signal y(n). The controllable amplifier 2802 is controlled by a thermal compressor module 2803 which is supplied with output signal y(n). As the attenuation from the thermal compressor module 2803 is applied upstream of the multi-band-compressor module 2801 (by way of controllable amplifier 2802), the thermal compressor module 2803 sets the basic volume level and the multi-band compressor module 2801 optimizes the sound based on the resulting frequency- dependent levels of the music signal.

[00125] However, the multi-band-compressor module 2801 may exhibit a considerable delay, so that gain reductions applied upstream of the multi-band-compressor module 2801 can take a relatively long time (e.g., >50ms) to become effective in the output signal y(t) and, thus, at the loudspeaker. Furthermore, loudspeakers are usually operated near their maximum SPL capabilities when the voice coil temperature is high. The peak limiter (e.g., multi-band compressor module 2801) may already apply considerable attenuation, which will be released when the thermal limiter (e.g., thermal limiter module 2803) in connection with controllable amplifier 2802 also provides for attenuation. The combination of both effects can result in response times of the thermal limitation that are too long for loudspeakers with short time constants for voice coil heating, particularly when operated by the multi-band-compressor module 2801, which allows for particularly high peak voltage.

[00126] This can be overcome with the structure described in connection with Figure 29. Gain reduction for thermal protection, which is controlled by a thermal section 2900, is performed downstream of the multi-band compressor 501 by way of a controllable amplifier 2901 with a controllable amplification TG. Gain reduction downstream of the multi-band compressor ensures that thermal limitation takes immediate effect without any delays. The same attenuation is applied within the sidebands of the multi-band compressor (e.g., by way of controllable amplifier 2902 with a controllable amplification TSGMB upstream of the band filters 505) in the fractional-band section 502 and, additionally, in the full-band section 503 by way of controllable amplifier 2903 with a controllable amplification TSGFB, which is connected between target filter 510 and compressor block 511, so that the multi-band compressor will still process the intermediate signal correctly. The controllable amplifiers 2901-2903 are controlled by a gain control block 2904, a thermal model block 2905 upstream of the gain control block 2904, and a power calculation block 2906 upstream of the thermal model block 2905. The blocks 2904-2906 are described below in connection with Figures 31- 36. The implementation shown in Figure 29 may result in excessive short term limitation by the peak limiter as the release of limitation may take some time, although this will normally not be noticed by the listener.

[00127] It should be noted that the variable gain element (controllable amplifier 2902) in the fractional-band section 502 may be alternatively placed upstream of the target (curve) filter 504. Another alternative is to dispose variable gain elements upstream of each compressor block 506, which would result in faster reaction of the fractional-band section to volume changes induced by the thermal limiter, while the fractional-band section would delay reaction with the phase shift of the target curve filter. Correspondingly, the variable-gain element 2903 in the full-band section 502 could be placed upstream of the target filter 510.

[00128] The approach outlined above can also be applied to implementing thermal limitation in a multi-band compressor with parallel structure as shown in Figure 30, which, in the present example, has only a single fractional-band but may also have multiple fractional- bands. The exemplary compressor shown in Figure 30 is based on the compressor shown in Figure 7, wherein a thermal section as described above in connection with Figure 29 is added. Furthermore, the same attenuation may be applied within the sidebands of the multi-band compressor, e.g., by way of controllable amplifier 3001 with a controllable amplification TSGMB, in the fractional-band section 602 and in the full-band section 603 by way of controllable amplifier 3002 with a controllable amplification TSGFB, which is connected between delay 611 and compressor block 612.

[00129] An exemplary way of calculating the instantaneous power p(n) applied to the loudspeaker is now described in connection with Figure 31. The output signal y(n) of a complete compressor module (not shown) is fed back into a thermal section (not shown) for power and subsequent temperature calculation. In a power calculation block 3000, which may be used as power calculation block 2906 in the compressor modules shown in Figures 29 and 30, the output signal y(n) may be first scaled to the voltage that is applied to the loudspeaker to be protected by way of an output voltage scaling (OVS) block 3101. As can be seen from Figure 32, a digital-analog converter 3201 may be connected between a compressor module 3200 with a thermal section and a power amplifier 3204-2 that has a gain G and drives a loudspeaker 3203. In order to scale the digital signal y(n) to the resulting voltage across the loudspeaker 3203, the OdBFs output voltage of the digital-analog converter 3201 and the gain G of the power amplifier 3201 are considered. The same applies for any other gain element connected between the compressor module 3200 and the loudspeaker 3203. Referring again to Figure 31, the voltage obtained from OVS block 3101 is then squared in a block 3102 and divided by the loudspeaker DC resistance R in a block 3103 to provide a signal p(n), representing the power consumption of the loudspeaker.

[00130] Alternatively, instead of the DC resistance R the minimum AC impedance within the audio range may be used, which includes eddy current losses within the voice coil former. Power calculation based on the DC resistance of the loudspeaker neglects the imaginary part of the complex loudspeaker impedance and can therefore lead to high deviations between the calculated and the actually consumed power of the loudspeaker. As can be seen from Figure 33, which depicts the impedance over frequency of an exemplary loudspeaker with 1.7 Ohms DC resistance within a broad range of the frequency spectrum, the impedance, which determines the current through the loudspeaker, is considerably higher than the DC resistance. [00131] In order to consider the influence of the complex impedance in the power calculation, the admittance function of the loudspeaker may be modeled by a filter which is part of the alternative power calculation block 3400 shown in Figure 34. For example, signal y(n) is scaled by an OVS block 3401 to the output voltage of the power amplifier and is subsequently filtered by an admittance filter 3402 with a transfer function HADM(z). As the output voltage scaling is usually implemented by multiplication with the appropriate factor, it may be included in the admittance filter 3402. The output of the admittance filter 3402 represents the current through the loudspeaker's voice coil. This is subsequently squared in a block 3403 and multiplied with the DC resistance R of the loudspeaker in a block 3404. Again, the minimum AC impedance within the audio range may be used instead of the DC resistance R as this will include eddy current losses within the voice coil former. The magnitude response of an exemplary admittance filter (e.g., applicable as filter 3402) derived from the admittance function shown in Figure 33 is illustrated in Figure 35. For example, the admittance filter 3402 may employ a multiple infinite impulse response (IIR) biquad structure or a finite impulse response (FIR) structure.

[00132] In order to model the temperature of the voice coil, a thermal model 3601 as shown in Figure 36 may be used. This model 3601 is applicable as model 2905 in the structure shown in Figures 29 and 30, and represents the heat transfer from the voice coil to the magnet and from the magnet to the ambience, e.g., the air within the loudspeaker enclosure. The model may also be extended to include further heat transfer to the air outside of the enclosure. In this model, the power loss p(n) within the loudspeakers voice coil, as input into the model, equals the heat transfer rate or heat flow q(n) over a series connection of two RC elements, which include a parallel connection of a thermal resistance and a thermal capacitance. One RC element includes a thermal capacitance Cth_vc and a thermal resistor Rth_vc and the other RC element includes a thermal capacitance Cth_M and a thermal resistor R±_M. The heat flow q(n) causes a temperature increase dt(n) over ambient temperature TA, likewise determining the voice coil temperature t(n).

[00133] Due to the thermal model's analogy to electrical circuits, Ohm's law as well as Kirchhoff's laws apply to this model. Therefore the following equations may be used for the calculation of the voice coil temperature:

-At

[00134] t_vc(ri) = t_{VC STEADY}(n) - (t_{VC STEADY}(n) - t_vc(n - 1)) * e^Tvc , [00135] t_vc STEADY (n) = t_M (n - 1) + q (n) * R_{th vc} ,

(

[00136] t_M (n) = t_{M STEADY}(n) - \ t_{M STEADY}(n) - t_M (n - 1) \ J * e—^TM ,

[00137] t_{M STEADY}(n) = T_A + q (n) * R_{th M} , wherein:

[00138] t_vc (n) = t(n) : Momentary voice coil temperature [°C],

[00139] t_vc (n— 1) : Voice coil temperature of the preceding calculation step [°C],

[00140] t_M (n) : Momentary magnet temperature [°C],

[00141] t_M (n— 1) : Magnet temperature of the preceding calculation step [°C]

[00142] At: Time increment between temperature calculations steps [s]

[00143] T_VC : Thermal time constant of the voice coil (Rth_vc*Cth_vc) [s]

[00144] Rth c- Thermal resistance from the voice coil to the magnet (pole tips) [K/W]

[00145] C_{th vc}: Thermal capacitance of the of the voice coil [Ws/K]

[00146] τ_Μ. Thermal time constant of the magnet (Rth_M*Cthjvi) [s]

[00147] R_th_M - Thermal resistance from the magnet to the environment [K/W]

[00148] C_{th M}: Thermal capacitance of the magnet [Ws/K]

[00149] q (n) : Momentary heat transfer rate (momentary power loss in the voice coil) [W]

[00150] T_A: Ambient temperature of the loudspeaker [°C]

[00151] Gain calculation as described above in connection with Figures 29 and 30 (gain control block 2904) may be realized as (digital) P, PI or PID controller. For example, to reduce audibility of gain changes, gain and time constants of PID sections may be different for rising and falling attenuations. Accordingly, gain reduction (attack phase of the thermal limiter) and gain increase (release phase) would exhibit different steepness of the gain ramp. Attack needs to be fast to avoid damage to the voice coil while release may be slow to minimize audibility of the gain change. The steepness of the gain change may be made dependent from the gradient of the voice coil temperature change or the distance from the thermal limit or a combination thereof. Furthermore, the steepness of the gain changes may be limited to avoid audible disturbances.

[00152] While the peak voltage limitation of a multi-band module can be adjusted to protect even multiple loudspeakers coupled to the amplifier with passive crossover networks, this is not the case for the thermal section as described above. This can be overcome with a structure shown in Figure 37, which is based on the structure shown in Figure 29. Instead of the thermal section 2900, a thermal section 3701 is used. In the thermal section 3701, the signal y(n) is split into a multiplicity of paths by way of crossover filters 3705. Each path includes a gain control block 3702, a thermal model block 3703 disposed upstream of the gain control block 3702, and power calculation block 3704 disposed upstream of the thermal model block 3703. The transfer functions of the passive crossover filters are approximated by the crossover filters 3705 which have transfer functions Ηχο_ι(ζ) to Ηχο__η(ζ). The crossover filters 3705 implement the actual transfer function that the passive crossover filters exhibit at the complex load formed by the loudspeaker impedance. With the resulting signal across the loudspeaker terminal's power calculation, thermal modeling and gain calculation are carried out individually for each loudspeaker using individual admittance functions and time constants for all loudspeakers. The tolerable overall gain gin(n), representing the minimum gain allowed by all loudspeakers, is then calculated as the minimum of all gains (e.g., by way of block 3706). With the structure shown in Figure 37, it is also possible to protect a single loudspeaker behind a passive crossover network (e.g. tweeter coupled to amp through a capacitor). In this case, only a single crossover filter and upstream gain calculation path are required in the thermal section 3701. It is also possible to implement the controlled gain distribution over multiple compressor channels, as described above in connection with Figures 23-25 for multiple multi-band compressor modules, by way of a multi-channel link module. In this case, the attenuation required in a single compressor channel due to thermal protection is fed into the multi-channel link module that compares the thermal attenuation for all channels in the same way as described for the gain of the full-band section. It is also possible to implement the controlled gain distribution over frequency, as described for the multi-band compressor module (e.g. module 501) with subsequent fractional-band section and full-band section, by means of a central gain distribution unit.

[00153] Referring to Figure 38, gain values determined in a signal analysis section 3801 are passed to a gain distribution unit 3802 that returns gains which are then applied in a gain application section 3803. In the structure shown in Figure 38, gain factors gREQ_B»i (n) to gREQ_Bn(n) are provided as outputs from comparator blocks 3804 in an analysis section 3801 and gain factors gAPP_Bi(n) to gAPP_Bn(n) are provided by way of coefficient calculation blocks 3807 to calculate filter coefficients for parametric filters 3806 in the main signal path of a gain application section 3803. Similarly, a local full-band gain gREQ_FB(n) forms an output signal from comparator block 3805 in the analysis section 3801 and a full-band gain g APP_FB (n) is used to control a variable gain block 3808 in the gain application section 3803, which provides output signal y(n) by way of a slope control block 3809. The variable gain block 3808 is connected in series with the parametric filters 3806 in the main signal path of the gain application section 3803. Slope control blocks 3810 may be connected upstream of the coefficient calculation blocks 3807. In the analysis section 3801, band filters 3811 are connected between a target filter 3813 and the parameter map blocks 3804, and a delay 3812 is connected between the target filter 3813 and the comparator block 3805. The target filter 3813 and a delay 3814 in the main signal path of the analysis section 3801, wherein the main signal path of the analysis section 3801 is disposed upstream of the main signal path of the gain application section 3803, are supplied with the input signal x(n).

[00154] In the signal analysis section 3801, the target filter 3813 with transfer function HMB_TAR(z) weights the (audio) input signal before it is split into n bands by the band filters 3811 with transfer functions HBF_l(z) to HBF_n(z). Then the signal is compared to a threshold CMP1 to CMPn in comparator blocks 3804 and the difference between the threshold and the input signal x(n) is output as a positive or negative gain value. There is also a branch for processing the full frequency spectrum, which includes the delay 3812 (delay time FB delay) instead of a band filter. However, the delay 3812 may be substituted by an all- pass filter. The all-pass filter can optionally be used to align the full-band signal with the band filtered signals that exhibit phase shift induced by the band filters. The full-band signal is also compared to a threshold FBCMP by way of comparator block 3805 and the difference between the threshold FBCMP and the signal is output as a positive or negative gain value.

[00155] In the gain application section 3803, slope control blocks 3810 with control functions SLP CTRLl to SLP CTRLn limit the rate of gain change that is effectively applied through the parametric filters 3806 with transfer functions HPAREQ_l(z) to HPAREQ_n(z) and the variable gain element 3808. The slope control blocks 3810 apply different slopes to increasing and decreasing gains. Their behavior is controlled by time constants that can be regarded as attack and release times. The function of the variable gain element 3808 can also be fully integrated in the parametric filters 3806 if these filters cover the whole frequency range of interest. Filter coefficients are calculated for all fractional-bands from the gain factors and applied to the parametric filters 3806 in the main signal path of the gain application section 3803.

[00156] An exemplary implementation of the gain distribution unit 3802 is shown in Figure 39. The required gain factors gREQ_Bi(n) to gREQ_Bn(n) of all fractional-bands are fed through mapping blocks 3901 with maps MAPI to MAPn, which map the gain factors gREQ_B 1 (n) to gREQ_Bn(n) to gain factors gAPP_Bi(n) to gAPP_Bn(n) in a linear or nonlinear fashion. Restriction of the frequency selective limitation is carried out with these parameter maps in the manner previously described in connection with the compressor modules with fractional-band sections and full-band sections. The resulting gain factors gAPP_Bi(n) to gAPP_Bn(n) are directly fed to the gain application section 3803.

[00157] For determining the full-band gain as applied to the variable gain element 3808 of the gain application section 3803, the minimum required gain of the gains gREQ_Bi(n) to gREQ_en(n) in all fractional-bands is evaluated, e.g., by a minimum detector block 3903, and then subtracted from the required full-band gain gREQ__FB(n), e.g., by way of a subtractor 3904. A compressor block 3905 with compressor function LEVI_FB limits a resulting gain factor gDEL_FB_FRB (n) to unity gain. In another branch, the required gain gREQ_Bi(n) to gREQ_Bn(n) in the fractional-bands is limited by compressors 3902 with compressor functions LIM1 to LEVln to unity gain and added to the inverted applied fractional-band gain factors gAPP_Bi(n) to gAPP_Bn(n), e.g., by way of adders 3906. The minimum gain factor gFB_MiN_REM_FRB (n) of all gain factors resulting thereof is detected, e.g., by way of minimum detector 3907, and added to gain factor gDEL_FB_FRB (n) , e.g., by way of adder 3908 which provides the full-band gain g APP_FB (n) .

[00158] As mentioned earlier, the variable gain element of the gain application section may be fully integrated into the parametric filters 3806 if these filters cover the whole frequency range of interest. An accordingly adapted structure may include an exemplary gain distribution unit 4001 is shown in Figure 40, in which the full range part of the applied gain is simply added to all fractional-band gains by way of adders 4002.

[00159] The combination of a fractional-band section and a full-band section with a controlled distribution of attenuations between both sections can be also applied in connection with feedback structures for determining the required limitation. Feedback structures can be applied if nonlinear models are used to determine a certain state variable of the loudspeaker. Linear models may also be used within feedback structures, but for linear models feedback structures are not required. In the exemplary structure shown in Figure 41, the required attenuations, and in the case of bass boost or loudness compensation also gains for a fractional-band control section 4101 and a full-band control section 4102, are calculated within respective feedback loops FRB LOOP1 to FRB LOOPn and FB LOOP. The loops FRB LOOP1 to FRB LOOPn and FB LOOP each comprise a modeling block 4103, 4113 which includes a nonlinear model NL Model for a certain loudspeaker state variable as, for example, membrane excursion. Furthermore, in each loop a controller 4104, 4114 which implements a control function EAMP or FBEAMP controls the gain in a manner that a target value for the state variable is met (e.g., maximum allowable membrane excursion). The gains in the fractional-bands are then recalculated as has been illustrated above for the feed-forward version of the multi-band compressor module to restrict the tonality change, e.g., by way of map blocks 4105 with transfer functions DSTMAP1 to DSTMAPn. Restrictions are again based on equal loudness curves or other suitable psychoacoustic models (e.g. model describing the masking effect of simultaneous tones). The respective restricted gain is applied to parametric filters 4106 in the full-band control section 4102 and to parametric filters 4110 in a main signal path 4107. A delay 4108 with a delay time FB delay serves to temporally separate the fractional-band control section 4101 from the full-band control section 4102 in order to avoid excessive limitation, which would occur if the fractional-band attenuation is not applied before the full-band control section 4102 analyzes the input signal x(n) by way of a full-band loop LOOP. An optional delay 4109 with a delay time MP delay, which is connected upstream of parametric filters 4110 with transfer functions HPAREQ_l(z) to HPAREQ_n(z) and a variable gain element 4111 in the main signal path 4107, serves to avoid overshoots for the complete compressor module. The main signal path outputs the output signal y(n).

[00160] Another variable gain element 4112 is connected downstream of the series- connected parametric filters 4106 and upstream of the modeling block 4113 to finally establish, in connection with the modeling block 4113 and the controller 4114, (control) loop FB LOOP. The controller 4114, which has the control function FBEAMP and which is connected downstream of the modeling block 4113, controls the variable gain element 4112 and, via a slope control block 4115 with a transfer function FBSLPCTRL, the variable gain element 4111. In the fractional-band control section 4101, slope control blocks 4116 with control functions SLPCTRLi to SLPCTRL_n are disposed upstream of the map blocks 4105 and downstream of controllers 4104. Map blocks 4117 with transfer functions BPAR MAPi to BPAR MAPn are also disposed downstream of the controllers 4104 to control series- connected parametric filters 4118, wherein the series-connection receives the input signal x(n) and feeds its output signal to band filters 4119. As can be seen, the (control) loops FRB LOOPl to FRB LOOPn for the parametric filters 4118 are established by the nonlinear models 4103, the band filters 4119, controllers 4104 and map blocks 4117. Furthermore, one or multiple parametric filters 4118 may be included in the respective feedback loops FRB LOOP 1 to FRB LOOP n. As multiple loops share common elements, these loops are effectively nested, resulting in mutual coupling between loops.

[00161] In order to avoid mutual coupling, the feedback loops of the fractional-band section may also be separated per parametric filter as shown in Figure 42. Instead of series- connection of the parametric filters a parallel structure is employed in which parametric filters 4201 are each supplied via a corresponding band filter 4202 with the input signal x(n). The parametric filters 4201 are controlled by map blocks 4203 which are disposed upstream of controllers 4204. The controllers 4204 are disposed downstream of modeling blocks 4205 and feed the slope control blocks 4116, similarly as in the structure shown in Figure 41. To finally establish the (control) loops FRB LOOPl to FRB LOOPn the modeling blocks 4205 are supplied with the output signals from parametric filters 4201.

[00162] Another option is to replace the parametric filters 4201 in the feedback loops of the fractional-band control section with variable gain elements 4301 as shown in Figure 43. In order to ensure quick tracking of the required attenuation, the feedback loops can in this case be run at multiple clock speed of the remaining signal flow. Furthermore, the map blocks 4203 may be substituted by direct connections.

[00163] In the fractional-band control section, parametric filters with transfer functions H_FRBEQ__I(Z) to H_FRBEQ_J(Z) are used to attenuate or, as the case may be, amplify the associated frequency band. Implementation options include IIR and FIR filters. The filters may be matched to the corresponding band filters with transfer functions H_BF__I(Z) to H_BFJ(Z) in the same loop if the limited frequency response is expected to be flat for a certain loudspeaker model (NL Model) with equal amplitude response over the complete frequency spectrum of interest. In this case, the parametric filters and band filters should be complementary in the sense that they add up to provide a flat response curve if configured with the same positive or negative gain. It should be noted that a flat limited frequency response is not mandatory for the application of the compressor and that some applications may explicitly need non-flat limited frequency responses for flat target curves.

[00164] Band filters HBF_l(z) to HBF_n(z) band- limit the frequency range processed by the respective control loops, resulting in n fractional-bands. It should be noted that n may also be 1 which means there is only a single fractional-band. The fractional-bands may not cover the full audio bandwidth. If the limited frequency response of the fractional-band compressor section is required to be flat for a loudspeaker model (NL Model) with equal amplitude response over the complete frequency spectrum of interest, the band filters and the corresponding parametric filters may be matched. Matching includes the center frequency and, largely, the bandwidth but also the complete shape of the filter curves. To avoid insufficient limitation (under-limitation) at the frequencies that fall between the filter bands it may be required to have slightly higher bandwidth or a filter curve shape that applies less attenuation at the sides of the bands for the band filters than for the filters in the main signal path. FIR filters or multiple IIR biquads can be designed to give the required transfer function, however it is also possible to obtain a sufficiently flat limited response with standard IIR band-pass filters.

[00165] The model NL Model in the fractional-band control section is a nonlinear model of a loudspeaker state variable as, for example, membrane excursion. It converts the incoming signal into a signal representing the respective state variable for the given audio signal.

[00166] Error amplifiers with transfer functions EAMP1 to EAMPn work with the signal delivered by the non-linear model representing a state variable of the loudspeaker (e.g. excursion). They may be implemented as difference amplifiers with fixed reference (e.g. max. allowable excursion) followed by a filter that implements some kind of controller (e.g. P, PI, PD, PID). Control may be bidirectional (e.g. attenuation and amplification) or unidirectional. The error amplifiers are configured to output a positive and/or negative gain factor.

[00167] Parameter maps with maps BPAR MAPI to BPAR MAPn convert the gain signal from the corresponding error amplifiers into parameters that define the transfer function of the parametric filters with transfer functions HFRBEQ_I(Z) to HFRBEQ_J(Z) in the fractional-band control section. Dependent on the implementation of the parametric filters, the parameter sets may include all or at least a subset of the IIR or FIR filter coefficients that define the parametric filters. The parameter maps may be implemented as simple look-up tables that choose a set of filter coefficients based on the incoming gain. A two-stage approach is also possible with a look-up table or some kind of linear or nonlinear function that maps incoming gain values to other gain values in the first stage. For example, negative gain values of a certain range may be mapped to positive gain values while negative values from another range are mapped to other negative values by nonlinear or piecewise linear functions. In the second stage the filter coefficients are calculated. The function of the second stage may also be carried out by an additional external coefficient calculation block if the gain values are to be evaluated or modified in after calculation.

[00168] Slope Controllers with control functions SLPCTRL1 to SLPCTRLn control the slope of the gain signal for increasing and decreasing gain. They may, for example, be implemented as averaging filters with different time constants for increasing and decreasing input signal. These time constants can be referred to as attack time and release time.

[00169] Distribution maps DSTMAP1 to DSTMAPn convert the slope controlled gain signal from the corresponding feedback loop of the fractional-band control section into parameters that define the transfer function H_PAREQ__I(Z) to H_PAREQ_J(Z) of the parametric filters in the full-band control section and also in the main signal path. Depending on the implementation of these filters, the parameter sets may include all or at least a subset of the IIR or FIR filter coefficients that define the parametric filters. The conversion may be implemented by way of simple look-up tables for choosing a set of filter coefficients based on the incoming averaged gain. A two-stage approach is also possible with a look-up table or some kind of linear or nonlinear function that maps incoming gain values to other gain values in the first stage. For example, negative gain values of a certain range may be mapped to positive gain values while negative values from another range are mapped to other negative values by nonlinear or piecewise linear functions. In the second stage the filter coefficients are calculated. The function of the second stage may also be carried out by an additional external coefficient calculation block if the gain values are to be evaluated or modified after calculation.

[00170] In the full-band control section, the delay block with delay time FB delay may optionally be substituted by an all-pass filter. It is used to temporally separate the full-band control section from the fractional-band. It compensates any delay caused in the band filters, the non-linear models and the error amplifiers of the fractional-bands, at least partly.

[00171] Parametric filters with transfer functions H_PAREQ__I(Z) to H_PAREQ_J(Z) are used to attenuate or in some cases amplify the associated frequency band. Implementation options include IIR and FIR filters. The filters are matched to the corresponding parametric filters H_FRBEQ__I(Z) to H_FRBEQ_J(Z) of the fractional-band section regarding their covered bandwidth and slope. For example, shelving or equalizing (e.g., peak, notch) filters can be used for the parametric filters that share the same corner or center frequency and quality. [00172] The block var. gain receives a control signal that sets the gain which is immediately applied.

[00173] Model NL Model in the full-band control section is a nonlinear model of a loudspeaker state variable, such as, for example, membrane excursion. It converts the incoming audio signal into a signal representing the respective state variable for the given audio signal.

[00174] Error amplifier with transfer function FBEAMP works with the signal delivered by the nonlinear model representing a state variable of the loudspeaker (e.g. excursion). It may be implemented as difference amplifiers with fixed reference (e.g. max. allowable excursion) followed by a filter that implements a controller (e.g. P, PI, PD, PID). Control may be bidirectional (e.g. attenuation and amplification) or unidirectional. The error amplifiers are configured to output a positive and/or negative gain factor.

[00175] Slope Controller with control function FBSLPCTRL controls the slope of the gain signal for increasing and decreasing gain. It may be implemented as an averaging filter with different time constants for increasing and decreasing input signal. These time constants can be referred to as attack time and release time.

[00176] In the main signal path, the delay with delay time MP delay can be used to avoid overshoots of the complete compressor module. It may compensate all delays caused inside the fractional-band control section as well as in the full-band control section. Parametric filters HPAREQ_I(Z) to HPAREQ_J(Z) are equivalent to the corresponding filters in the full-band control section. The variable gain applies the gain values from the slope control in the full- band control section to the audio signal in the main signal path.

[00177] The multi-band compressors described above allow high perceived loudness while minimizing the audible disturbances caused by the limitation process. By applying continuously variable effective compressor thresholds over the complete frequency range, SPL reduction is minimized to the real limitations of the loudspeaker. Separate analysis and limitation of several frequency bands reduces mutual interaction between these bands thereby increasing perceived loudness and reducing pumping effects. Furthermore, the restriction of compressor induced tonality changes is based on human loudness perception models. Time constants for gain changes can be optimized for the processed frequency range thereby reducing uncorrelated noise. Through use of side signal paths for signal analysis, the audio signal is not processed if no limitation is required. Synchronization between multiple compressor modules allows coherent limitation of multi-channel audio systems that preserve the tonal balance and spatial representation. Signal overshoots can be avoided, thus providing reliable loudspeaker protection. Additionally, the compressor structure may be used for dynamic bass boost and loudness compensation, which are only applied if the loudspeaker capabilities allow for it. In combination with a generator of higher harmonics, virtual bass can be added to the signal, augmenting or replacing only the low frequency content that has been attenuated by the limiter. If not otherwise noted, the multi-band compressor described above can be applied as peak voltage or loudspeaker excursion limiter. However, it may also be adapted to limit peak or RMS current or power applied to loudspeakers or the voice coil temperature of a loudspeaker.

[00178] While various embodiments of the invention have been described, it will be apparent to those of ordinary skill in the art that many more embodiments and implementations are possible within the scope of the invention. Accordingly, the invention is not to be restricted except in light of the attached claims and their equivalents.

Claims

1. An audio signal compressor configured to compress by amplitude an input audio signal within an audio signal frequency band to provide an output audio signal, the compressor comprising at least two compression frequency bands in which the audio signal is compressed, at least one of the compression frequency bands being narrower than the complete audio signal frequency band.

2. The compressor of claim 1, where compression is based on a compression function, the compression function being frequency dependent and/or amplitude dependent.

3. The compressor of claim 1 or 2, further configured to evaluate the level or amplitude of the input audio signal within at least one analysis frequency band to be taken as a basis for determination of a gain or attenuation for compressing the input audio signal applied within at least one of the compression frequency bands.

4. The compressor of claim 3, where at least one of the at least one analysis frequency band and at least one of the at least two compression frequency bands comprise different bandwidths; and where the gain or attenuation applied in the at least one of the at least two compression frequency bands is based on the result of the evaluation carried out in the at least one of the at least one analysis frequency band.

5. The compressor of claim 3 or 4, further configured to evaluate the level or amplitude of the input audio signal within at least one of the at least one analysis frequency band based on at least one weighting function.

6. The compressor of claim 5, where at least one of the at least one weighting function is frequency dependent within at least one of the at least one analysis frequency band.

7. The compressor of any of claims 2-6, where the compression function within at least one of the compression frequency bands is based on at least one of a group comprising a human audio perception model, at least one characteristic of a loudspeaker, at least one characteristic of a passive component connected to the loudspeaker, at least one characteristic of an audio playback device and at least one characteristic of an audio system.

8. The compressor of claim 6 or 7, where at least one of the at least one frequency- dependent weighting function is based on at least one of a group comprising at least one characteristic of a loudspeaker, at least one characteristic of a passive component connected to the loudspeaker, at least one characteristic of an audio playback device, at least one characteristic of an audio system, and at least one characteristic of a human audio perception model.

9. The compressor of claim 7 or 8, where the at least one characteristic of a loudspeaker and/or the at least one characteristic of a passive component connected to the loudspeaker includes at least one of a magnitude vs. frequency characteristic of the loudspeaker and/or passive component, an admittance vs. frequency characteristic of the loudspeaker and/or passive component, a maximum withstand voltage vs. frequency of the loudspeaker and a maximum voltage vs. frequency at which acoustic artifacts are tolerable.

10. The compressor of any of claims 7-9, where the at least one characteristic of an audio playback device and/or the at least one characteristic of an audio system includes at least one of a magnitude vs. frequency characteristic of the audio playback device and/or audio system, a maximum withstand voltage vs. frequency of the audio playback device and/or audio system, and a maximum voltage vs. frequency at which acoustic artifacts are tolerable.

11. The compressor of any of claims 2-10, where the compression function in at least one compression frequency band is partly determined by at least one level-dependent control function.

12. The compressor of claim 11, where the at least one level-dependent control function is based on at least one human audio perception model.

13. The compressor of any of claims 1-12, further comprising at least one frequency specific gain element configured to provide the compression in at least one compression frequency band.

14. The compressor of claim 13, where the at least one frequency specific gain element comprises at least one controllable filter with a dynamically controllable transfer function, the transfer function being configured to provide the compression in at least one compression frequency band.

15. The compressor of claim 14, wherein the transfer function of at least one of the at least one controllable filter applied to compress the audio signal is at least based on a human audio perception model.

16. The compressor of any of claims 1-15, further configured to limit voltage, current or power applied to a loudspeaker.

17. The compressor of any of claims 1-16, further comprising a main signal path and at least one side signal path, the main signal path being configured to amplify or attenuate the input audio signal with at least one controllable gain and/or at least one controllable transfer function within the compression frequency bands, and the at least one side signal path being configured to evaluate the input audio signal and to control the gain and/or at least one controllable transfer function in the main signal path dependent on results of the evaluation.

18. The compressor of any of claims 1-17, further comprising at least two compression sections that compress the input audio signal or at least one intermediate audio signal within at least one compression frequency band per compression section to provide at least one intermediate audio signal or the output audio signal, where each compression section is chronologically separated from at least one other compression section with regard to the input audio signal or the at least one intermediate audio signal so that the compression of one of any two separated sections is at least partially applied to the input audio signal or the at least one intermediate audio signal before the other section evaluates the input audio signal or the at least one intermediate audio signal.

19. The compressor of claim 18, configured to partially or completely control the distribution of compression sufficient to reach a first limitation or compression target between at least two compression sections or at least two compression bands with at least one of:

configuration of at least one of the at least one frequency-dependent weighting functions;

configuration of at least one of the at least one level-dependent control function;

configuration of reference amplitude or reference levels for analysis of the input audio signal or intermediate audio signal;

control of overshoots of the intermediate signal with respect to a second compression target;

configuration of the transfer function of at least one of the at least one controllable filter applied to compress the audio signal;

an amount of chronological separation between separated sections.

20. The compressor of claim 18 or 19, where at least one of the compression sections is a fractional compression section configured to receive the input audio signal or one of the at least one intermediate audio signal, and to compress the respective audio signal within at least one fractional band of the audio frequency band to provide one of the at least one intermediate audio signal, the at least one fractional band being narrower than the complete audio signal frequency band.

21. The compressor of claim 20, where at least one of the at least two compression sections is a full-band compression section configured to receive one of the at least one intermediate audio signal and to compress the respective audio signal within the full audio frequency band to provide the output audio signal or one of the at least one intermediate audio signal.

22. The compressor of claim 20 or 21, where at least one of the fractional compression sections is configured to control the compression of the input audio signal or the compression of one of the at least one intermediate audio signal within the at least one fractional band of the audio frequency band based on at least one first weighting function and/or at least one first control function, and the at least one of the at least one full-band compression section is configured to control the compression of one of the at least one intermediate audio signal within the full audio frequency band based on at least one second weighting function and/or at least one second control function.

23. The compressor of claim 22, where the first weighting function and the second weighting function are identical.

24. The compressor of any of claims 21-23, where

at least one of the at least one fractional compression section and the at least one full- band compression section comprises a main signal path and at least one side signal path; the main signal path is configured to amplify or attenuate the input audio signal or intermediate audio signal with controllable gain and/or controllable transfer function the input audio signal within the at least one fractional-band or the full audio signal frequency band; and

the at least one side signal path is configured to evaluate the input audio signal or the intermediate audio signal based on at least one of the first weighting function, second weighting function, first control function, and second control function, and to control the gain of the main signal path dependent on results of the evaluation.

25. The compressor of claim 24, comprising

one fractional compression section and one full-band compression section that each comprise a main signal path and at least one side signal path; where

the main signal path of the fractional compression section is configured to amplify or attenuate the input audio signal with a controllable gain and/or controllable transfer function within the at least one fractional-band to output the intermediate signal;

the main signal path of the full-band compression section is configured to amplify or attenuate the intermediate audio signal with a controllable gain within the full audio signal frequency band; and the at least one side signal path of the fractional compression section is configured to evaluate the input audio signal based on at least one of the first weighting function and/or first control function, and to control the gain of the main signal path of the fractional compression section within at least one fractional band dependent on results of the evaluation.

26. The compressor of claim 25, where the at least one side signal path of the full-band compression section is configured to evaluate the intermediate audio signal based on at least one of the second weighting function and the second control function, and to control the gain of the main signal path of the full-band compression section dependent on results of the evaluation.

27. The compressor of claim 25, where

the first weighting function and the second weighting function are identical;

one side signal path or a common part of multiple nested side signal paths of the fractional compression section is configured to apply frequency-dependent weighting to the input audio signal as defined by the first weighting function to generate a weighted input audio signal;

all side signal paths of the fractional compression section are configured to provide a representation of the gain or attenuation applied in the main signal path of the fractional compression section within the at least one fractional band to the at least one side signal path of the full-band compression section;

the at least one side signal path of the full-band compression section is configured to receive the representation of the gain or attenuation applied in the main signal path of the fractional compression section within the at least one fractional band and the weighted input audio signal and to amplify or attenuate the weighted input audio signal with a controllable gain and/or controllable transfer function within the at least one fractional-band of the fractional compression section in accordance with the gain or attenuation received from the respective side signal path of the fractional compression section to receive a compressed weighted input audio signal, where the compression applied to the weighted input audio signal in the side path of the full-band compression section is identical to the compression applied to the input audio signal in the main path of the fractional-band compression section; and

the at least one side path of the full-band compression section is further configured to evaluate the compressed weighted input audio signal based on at least the second control function and to control the gain of the main signal path of the full-band compression section dependent on results of the evaluation.

28. The compressor of any of claim 18-27, where at least one of the main signal path in the fractional-band compression section and the main signal path in the full-band compression section comprises at least one time delay element configured to delay the input audio signal or the intermediate audio signal to compensate at least partly for a time delay occurring during signal evaluation in the respective side signal path or paths, and/or to compensate for ramping of gain and/or filter parameters within at least one main signal path.

29. The compressor of claim 27 or 28, further comprising at least one time delay in the at least one side signal path of the full-band compression section configured to delay the weighted input audio signal to compensate at least partly for a time delay occurring during signal evaluation in the side signal path of the fractional-band compression section and/or ramping of gain and/or filter parameters within the side signal path of the full-band compression section.

30. The compressor of any of claims 20-29, where the main signal path of the fractional-band compression section comprises at least one controllable filter configured to apply gain or attenuation within at least one compression frequency band, and where the filter is controlled by the at least one side signal path.

31. The compressor of claim 30, further comprising a multiplicity of side signal paths in the fractional-band compression section, where the main signal path of the fractional-band compression section comprises a multiplicity of series-connected controllable filters that are controlled by the multiplicity of side signal paths in the fractional-band compression section.

32. The compressor of any of claims 22-31, where at least one of first weighting function, second weighting function, first control function, and second control function is based on a human audio perception model.

33. The compressor of claim 32, where the human audio perception model includes at least one of equal loudness curves, frequency groups and masking effect.

34. The compressor of any of claims 22-33, where at least one of first weighting function and second weighting function is based on loudspeaker or audio system characteristics.

35. The compressor of claim 34, further configured to limit power, voltage or current supplied to the loudspeaker.

36. The compressors of any of claims 1-35, further configured to control supplementation or substitution of components of the input audio signal by supplementation or substitution signals that are higher harmonics of the corresponding components of the input audio signal.

37. The compressors of claim 36, further configured to control the supplementation or complete substitution of components of the input audio signal by supplementation signals that are higher harmonics of the corresponding components of the input audio signal by way of the control of the amplitude and/or the frequency range of components of the input audio signal that are used as fundamental signals to generate the higher harmonics.

38. The compressor of any of claims 1-17, comprising:

a multitude of side signal paths configured to evaluate the level or amplitude of the input audio signal within a multitude of analysis frequency bands based on at least one frequency-dependent weighting function;

a multitude of controllable filters or gains configured to receive the input audio signal, to compress the audio signal within multiple bands and to output the compressed audio signal; and

a central gain distribution unit configured to receive the evaluation result of the multitude of side signal paths and to conduct a collective evaluation on all evaluation results of the multitude of side signal paths and based on the results of that collective evaluation control the compression applied by the multitude of controllable filters or gains.

39. The compressor of any of claims 1-38, further configured to at least partly compensate changes of relative loudness of spectral components of the audio signal as perceived by human listeners for varying playback levels, over parts of the audio signal frequency range or the complete audio signal frequency range.

40. The compressor of any of claims 1-39, where the compression is frequency- specifically controlled so that an extension to lower frequencies or the level of the magnitude vs. frequency characteristic of a loudspeaker and/or audio system is increased or decreased for certain parts of the loudspeaker's dynamic range or the audio system's dynamic range.

41. The compressors of any of claims 1-40, where the compressor comprises a downstream gain input configured to receive a representation of any momentary gain applied to the audio signal downstream of the compressor by any kind and number of dynamically changing gain stages; and the compressor is configured to adapt the compression applied to the audio signal in approximately the same way as would be the case for any gain changes occurring upstream of the compressor.

42. The compressor of claim 41, comprising a controllable side signal path compensation gain element within at least one side signal path configured to apply the gain received through the downstream gain input within the signal evaluation of the side signal path.

43. The compressor of any of claims 1-42, further comprising a thermal limiter section configured to evaluate the temperature of at least one loudspeaker and/or at least one audio system component and to reduce the level or amplitude of the audio signal supplied to the respective loudspeaker and/or audio system component so that the temperature is kept below a certain threshold, wherein the thermal limiter section is configured to receive the output audio signal and to control the level of the input audio signal.

44. The compressor of any of claims 1-42, further comprising a thermal limiter section configured to evaluate the temperature of at least one loudspeaker and/or at least one audio system component and to reduce the level or amplitude of the audio signal supplied to the respective loudspeaker and/or audio system component so that the temperature is kept below a certain threshold; the thermal limiter section configured to receive the output audio signal and to control the level of the output audio signal.

45. The compressor of claim 44, further configured to supply a representation of the momentary gain that is applied to the output audio signal to the at least one side signal path compensation gain element.

46. The compressor of claim 43 or 44, where the audio system component is an electronic crossover component of a passive crossover network of a loudspeaker system comprising a single loudspeaker or multiplicity of loudspeakers.

47. The compressor of any of claims 43-46, where the thermal limiter section is configured to evaluate the power applied to a loudspeaker and/or audio system component based on at least one filter derived from the admittance vs. frequency characteristic of at least one loudspeaker and/or audio system component.

48. The compressor of any of claims 43-47, where the thermal limiter section is configured to evaluate the power applied to a loudspeaker and/or audio system component based on at least one filter that approximates the transfer function of at least a part of the loudspeaker and/or audio system component.

49. The compressor of any of claims 43-48, where the thermal limiter section is configured to evaluate the temperature of the at least one loudspeaker and/or the at least one audio system component based on at least one thermal model.

50. A compressor arrangement comprising a multiplicity of compressors according to claims 1-49, where the multiplicity of compressors are connected to a multiplicity of audio channels; where the multitude of audio signal compressors is connected with each other via a multichannel link module configured to coherently control the compression in the multiplicity of channels;

the multi-channel link module is configured to receive signal evaluation results for at least one frequency band in at least two of the multiplicity of audio signal compressors and to control the gain of the multiplicity of audio signal compressors based on a comparison of the signal evaluation results.

51. An audio signal compressing method configured to compress by amplitude an input audio signal within an audio signal frequency band to provide an output audio signal, the method comprising at least two compression frequency bands in which the audio signal is compressed, at least one of the compression frequency bands being narrower than the complete audio signal frequency band.

52. The method of claim 51, where compression is based on a compression function, the compression function being frequency dependent and/or amplitude dependent.

53. The method of claim 51 or 52, further configured to evaluate the level or amplitude of the input audio signal within at least one analysis frequency band to be taken as a basis for determination of compression gain or attenuation applied within at least one of the compression frequency bands to compress the input audio signal.

54. The method of claim 53, where at least one of the at least one analysis frequency band and at least one of the at least two compression frequency bands comprise different bandwidths; and where the gain or attenuation applied in the at least one of the at least two compression frequency bands is based on the result of the evaluation carried out in the at least one of the at least one analysis frequency band.

55. The method of claim 53 or 54, further configured to evaluate the level or amplitude of the input audio signal within at least one of the at least one analysis frequency band based on at least one weighting function.

56. The method of claim 55, where at least one of the at least one weighting function is frequency dependent within at least one of the at least one analysis frequency band.

57. The method of any of claims 52-56, where the compression function within at least one of the compression frequency bands is based on at least one of a group comprising: a human audio perception model, at least one characteristic of a loudspeaker, at least one characteristic of a passive component connected to the loudspeaker, at least one characteristic of an audio playback device and at least one characteristic of an audio system.

58. The method of claim 56 or 57, where at least one of the at least one frequency- dependent weighting function is based on at least one of a group comprising at least one characteristic of a loudspeaker, at least one characteristic of a passive component connected to the loudspeaker, at least one characteristic of an audio playback device, at least one characteristic of an audio system, and at least one characteristic of a human audio perception model.

59. The method of claims 57 or 58, where the at least one characteristic of a loudspeaker and/or the at least one characteristic of a passive component connected to the loudspeaker includes at least one of a magnitude vs. frequency characteristic of the loudspeaker and/or passive component, an admittance vs. frequency characteristic of the loudspeaker and/or passive component, a maximum withstand voltage vs. frequency of the loudspeaker and a maximum voltage vs. frequency at which acoustic artifacts are tolerable.

60. The method of any of claims 57-59, where the at least one characteristic of a loudspeaker and/or the at least one characteristic of a passive component connected to the loudspeaker includes at least one of a magnitude vs. frequency characteristic of the loudspeaker and/or passive component, an admittance vs. frequency characteristic of the loudspeaker and/or passive component, a maximum withstand voltage vs. frequency of the loudspeaker and a maximum voltage vs. frequency at which acoustic artifacts are tolerable.

61. The method of any of claims 52-60, where the compression function in at least one compression frequency band is partly determined by at least one level-dependent control function.

62. The method of claim 61, where the at least one level-dependent control function is based on at least one human audio perception model.

63. The method of any of claims 51-62, further comprising at least one frequency specific gain configured to provide the compression in at least one compression frequency band.

64. The method of claim 63, where the at least one frequency specific gain comprises at least one controllable filtering with a dynamically controllable transfer function to provide the compression in at least one compression frequency band.

65. The method of claim 64, wherein the transfer function of at least one controllable filtering configured to compress the audio signal is at least based on a human audio perception model.

66. The method of any of claims 51-65, further configured to limit voltage, current or power applied to a loudspeaker.

67. The method of any of claims 51-66, further comprising a main signal path and at least one side signal path, the main signal path being configured to amplify or attenuate the input audio signal with at least one controllable gain and/or at least one controllable transfer function within the compression frequency bands, and the at least one side signal path being configured to evaluate the input audio signal and to control the gain or attenuation of the controllable gain and/or at least one controllable transfer function in the main signal path dependent on results of the evaluation.

68. The method of any of claims 51-67, further comprising at least two compression procedures that compress the input audio signal or at least one intermediate audio signal within at least one compression frequency band per compression procedure to provide at least one intermediate audio signal or the output audio signal, where each compression procedure is chronologically separated from at least one other compression procedure with regard to the input audio signal or the at least one intermediate audio signal so that the compression of one of any two separated procedures is at least partially applied to the input audio signal or the at least one intermediate audio signal before the other procedure evaluates the audio signal or the at least one intermediate audio signal.

69. The method of claim 68, configured to partially or completely control the distribution of compression sufficient to reach a first limitation or compression target between at least two compression procedures or at least two compression bands with at least one of:

configuration of at least one of the at least one level-dependent control functions;

configuration of the transfer function of at least one of the at least one controllable filters applied to compress the audio signal;

an amount of chronological separation between separated sections.

70. The method of claim 68 or 69, where at least one of the compression procedures is a fractional compression procedure configured to receive the input audio signal or one of the at least one intermediate audio signal, and to compress the respective audio signal within at least one fractional band of the audio frequency band to provide one of the at least one intermediate audio signal, the at least one fractional band being narrower than the complete audio signal frequency band.

71. The method of claim 70, where at least one of the at least two compression procedures is a full-band compression procedure configured to receive one of the at least one intermediate audio signal and to compress the respective audio signal within the full audio frequency band to provide the output audio signal or one of the at least one intermediate audio signal.

72. The method of claim 70 or 71, where at least one of the fractional compression procedures is configured to control the compression of the input audio signal within the at least one fractional band of the audio frequency band based on at least one first weighting function and/or at least one first control function, and the full-band compression procedure is configured to control the compression of the intermediate audio signal within the full audio frequency band based on at least one second weighting function and/or at least one second control function.

73. The method of claim 72, where the first weighting function and the second weighting function are identical.

74. The method of any of claims 71-73, where

at least one of the at least one fractional compression procedure and the at least one full-band compression procedure comprises a main signal path and at least one side signal path;

the main signal path is configured to amplify or attenuate the input audio signal or intermediate audio signal with controllable gain and/or controllable transfer function within the at least one fractional-band or the full audio signal frequency band; and

75. The method of claim 74, comprising:

one fractional compression procedure and one full-band compression procedure that each comprise a main signal path and at least one side signal path; where

the main signal path of the fractional compression procedure is configured to amplify or attenuate the input audio signal with a controllable gain and/or controllable transfer function within the at least one fractional-band to output the intermediate signal;

the main signal path of the full-band compression procedure is configured to amplify or attenuate the intermediate audio signal with a controllable gain within the full audio signal frequency band; and

the at least one side signal path of the fractional compression procedure is configured to evaluate the input audio signal based on at least one of the first weighting function and/or first control function, and to control the gain of the main signal path of the fractional compression procedure within at least one fractional band dependent on results of the evaluation.

76. The method of claim 75, where the at least one side signal path of the full-band compression procedure is configured to evaluate the intermediate audio signal based on at least one of the second weighting function and the second control function, and to control the gain of the main signal path of the full-band compression procedure dependent on results of the evaluation.

77. The method of claim 75, where

the first weighting function and the second weighting function are identical;

one side signal path or a common part of multiple nested side signal paths of the fractional compression procedure is configured to apply frequency-dependent weighting to the input audio signal as defined by the first weighting function to generate a weighted input audio signal; all side signal paths of the fractional compression procedure are configured to provide a representation of the gain or attenuation applied in the main signal path of the fractional compression procedure within the at least one fractional band to the at least one side signal path of the full-band compression procedure;

the at least one side signal path of the full-band compression procedure is configured to receive the representation of the gain or attenuation applied in the main signal path of the fractional compression section within the at least one fractional band and the weighted input audio signal and to amplify or attenuate the weighted input audio signal with a controllable gain and/or controllable transfer function within the at least one fractional-band of the fractional compression procedure in accordance with the gain or attenuation received from the respective side signal path of the fractional compression procedure to receive a compressed weighted input audio signal, where the compression applied to the weighted input audio signal in the side path of the full-band compression procedure is identical to the compression applied to the input audio signal in the main path of the fractional-band compression procedure; and

the at least one side path of the full-band compression procedure is further configured to evaluate the compressed weighted input audio signal based on at least the second control function and to control the gain of the main signal path of the full-band compression procedure dependent on results of the evaluation.

78. The method of any of claim 68-77, where at least one of the main signal path in the fractional-band compression procedure and the main signal path in the full-band compression procedure comprises at least one time delay configured to delay the input audio signal or the intermediate audio signal to compensate at least partly for a time delay occurring during signal evaluation in the respective side signal path or paths, and/or to compensate for ramping of gain and/or filter parameters within at least one main signal path.

79. The method of claim 77 or 78, further comprising at least one time delay in the at least one side signal path of the full-band compression section configured to delay the weighted input audio signal to compensate at least partly for a time delay occurring during signal evaluation in the side signal path of the fractional-band procedure and/or ramping of gain and/or filter parameters within the side signal path of the full-band compression procedure.

80. The method of any of claims 70-79, where the main signal path of the fractional- band compression procedure comprises at least one controllable filter configured to apply gain or attenuation within at least one compression frequency band, the filter is controlled by the at least one side signal path.

81. The method of claim 80, further comprising a multiplicity of side signal paths in the fractional-band compression procedure, where the main signal path of the fractional-band compression procedure comprises a multiplicity of series-connected controllable filters that are controlled by the multiplicity of side signal paths in the fractional-band compression procedure.

82. The method of any of claims 72-81, where at least one of first weighting function, second weighting function, first control function, and second control function is based on a human audio perception model.

83. The method of claim 82, where the human audio perception model includes at least one of equal loudness curves, frequency groups and masking effect.

84. The method of any of claims 72-33, where at least one of first weighting function and second weighting function is based on loudspeaker or audio system characteristics.

85. The method of claim 84, further configured to limit power, voltage or current supplied to the loudspeaker.

86. The methods of any of claims 51-85, further configured to control supplementation or substitution of components of the input audio signal by supplementation or substitution signals that are higher harmonics of the corresponding components of the input audio signal.

87. The methods of claim 86, further configured to control the supplementation or complete substitution of components of the input audio signal by supplementation signals that are higher harmonics of the corresponding components of the input audio signal by way of the control of the amplitude and/or the frequency range of components of the input audio signal that are used as fundamental signals to generate the higher harmonics.

88. The method of any of claims 51-67, comprising:

89. The method of any of claims 51-88, further configured to at least partly compensate changes of relative loudness of spectral components of the audio signal as perceived by human listeners for varying playback levels, over parts of the audio signal frequency range or the complete audio signal frequency range.

90. The method of any of claims 51-89, where the compression is frequency- specifically controlled so that an extension to lower frequencies or the level of the magnitude vs. frequency characteristic of a loudspeaker and/or audio system is increased or decreased for certain parts of the loudspeaker's dynamic range or the audio system's dynamic range.

91. The methods of any of claims 51-90, where the method comprises a downstream gain input configured to receive a representation of any momentary gain applied to the audio signal downstream of the method by any kind and number of dynamically changing gain stages; and the method is configured to adapt the compression applied to the audio signal in approximately the same way as would be the case for any gain changes occurring upstream of the method.

92. The method of claim 91, comprising a controllable side signal path compensation gain within at least one side signal path configured to apply the gain received through the downstream gain input within the signal evaluation of the side signal path.

93. The method of any of claims 51-92, further comprising a thermal limiter procedure configured to evaluate the temperature of at least one loudspeaker and/or at least one audio system component and to reduce the level or amplitude of the audio signal supplied to the respective loudspeaker and/or audio system component so that the temperature is kept below a certain threshold, wherein the thermal limiter section is configured to receive the output audio signal and to control the level of the input audio signal.

94. The method of any of claims 51-93, further comprising a thermal limiter procedure configured to evaluate the temperature of a loudspeaker and/or audio system component and to reduce the level or amplitude of the audio signal supplied to the respective loudspeaker and/or audio system component, so that the temperature is kept below a certain threshold; the thermal limiter procedure configured to receive the output audio signal and to control the level of the output audio signal.

95. The method of claim 94, further configured to supply a representation of the momentary gain that is applied to the output audio signal to the at least one side signal path compensation gain.

96. The method of claim 93 or 94, where the audio system component is an electronic crossover component of a passive crossover network of a loudspeaker system comprising a single loudspeaker or multiplicity of loudspeakers.

97. The method of any of claims 93-96, where the thermal limiter procedure is configured to evaluate the power applied to a loudspeaker and/or audio system component based on at least one filter derived from the admittance vs. frequency characteristic of at least one loudspeaker and/or audio system component.

98. The method of any of claims 93-97, where the thermal limiter procedure is configured to evaluate the power applied to a loudspeaker and/or audio system component based on at least one filter that approximates the transfer function of at least a part of the loudspeaker and/or audio system component.

99. The method of any of claims 93-98, where the thermal limiter procedure is configured to evaluate the temperature of the at least one loudspeaker and/or the at least one audio system component based on at least one thermal model.

100. A computer program code configured to perform the method of any of claims 51-

99.